Compare commits

...

35 Commits

Author SHA1 Message Date
Jedrzej Kosinski
50603859ab Merge branch 'master' into v3-definition 2025-06-01 01:51:04 -07:00
Benjamin Lu
180db6753f Add Help Menu in NodeLibrarySidebarTab (#8179) 2025-06-01 04:32:32 -04:00
Christian Byrne
d062fcc5c0 [feat] Add ImageStitch node for concatenating images (#8369)
* [feat] Add ImageStitch node for concatenating images with borders

Add ImageStitch node that concatenates images in four directions with optional borders and intelligent size handling. Features include optional second image input, configurable borders with color selection, automatic batch size matching, and dimension alignment via padding or resizing.

Upstreamed from https://github.com/kijai/ComfyUI-KJNodes with enhancements for better error handling and comprehensive test coverage.

* [fix] Fix CI issues with CUDA dependencies and linting

- Mock CUDA-dependent modules in tests to avoid CI failures on CPU-only runners
- Fix ruff linting issues for code style compliance

* [fix] Improve CI compatibility by mocking nodes module import

Prevent CUDA initialization chain by mocking the nodes module at import time,
which is cleaner than deep mocking of CUDA-specific functions.

* [refactor] Clean up ImageStitch tests

- Remove unnecessary sys.path manipulation (pythonpath set in pytest.ini)
- Remove metadata tests that test framework internals rather than functionality
- Rename complex scenario test to be more descriptive of what it tests

* [refactor] Rename 'border' to 'spacing' for semantic accuracy

- Change border_width/border_color to spacing_width/spacing_color in API
- Update all tests to use spacing terminology
- Update comments and variable names throughout
- More accurately describes the gap/separator between images
2025-06-01 04:28:52 -04:00
Jedrzej Kosinski
0d185b721f Created and handled NodeOutput class to be the return value of v3 nodes' execute function 2025-06-01 01:08:07 -07:00
filtered
456abad834 Update frontend to 1.21 (#8366) 2025-06-01 01:10:04 -04:00
comfyanonymous
19e45e9b0e Make it easier to pass lists of tensors to models. (#8358) 2025-05-31 20:00:20 -04:00
Jedrzej Kosinski
8642757971 Made V3 NODES_LIST work properly 2025-05-31 15:32:11 -07:00
kosinkadink1@gmail.com
de86d8e32b Attempting to simplify node list definition in a python file via NODES_LIST 2025-05-31 15:24:37 -07:00
kosinkadink1@gmail.com
8b331c5ca2 Made proper None checks in V1 translation class properties for ComfyNodeV3 2025-05-31 04:14:01 -07:00
Jedrzej Kosinski
937d2d5325 Fixed 'display' serialization for Float/IntergerInput, some commented out code made during exploration 2025-05-31 04:00:03 -07:00
ComfyUI Wiki
97f23b81f3 Bump template to 0.1.23 (#8353)
Correct some error settings in VACE
2025-05-30 23:05:42 -07:00
drhead
08b7cc7506 use fused multiply-add pointwise ops in chroma (#8279) 2025-05-30 18:09:54 -04:00
BennyKok
6c319cbb4e fix: custom comfy-api-base works with subpath (#8332) 2025-05-30 17:51:28 -04:00
Chenlei Hu
df1aebe52e Remove huchenlei from CODEOWNERS (#8350) 2025-05-30 17:27:52 -04:00
comfyanonymous
704fc78854 Put ROCm version in tuple to make it easier to enable stuff based on it. (#8348) 2025-05-30 15:41:02 -04:00
JettHu
1d9fee79fd Add node for regex replace(sub) operation (#8340)
* Add node for regex replace(sub) operation

* Apply suggestions from code review

add tooltips

Co-authored-by: Christian Byrne <abolkonsky.rem@gmail.com>

* Fix indentation

---------

Co-authored-by: Christian Byrne <abolkonsky.rem@gmail.com>
2025-05-30 15:08:59 -04:00
Jedrzej Kosinski
0400497d5e Merge branch 'master' into v3-definition 2025-05-30 02:49:02 -07:00
Jedrzej Kosinski
aeba0b3a26 Reduce code duplication for [pro] and [max], rename Pro and Max to [pro] and [max] to be consistent with other BFL nodes, make default seed for Kontext nodes be 1234. since 0 is interpreted by API as 'choose random seed' (#8337) 2025-05-29 17:14:27 -04:00
comfyanonymous
094306b626 ComfyUI version 0.3.39 2025-05-29 14:26:39 -04:00
filtered
31260f0275 Update templates 0.1.22 (#8334) 2025-05-30 03:52:27 +10:00
Robin Huang
f1c9ca816a Add BFL Kontext API Nodes. (#8333)
* Added initial Flux.1 Kontext Pro Image node - recreated branch to save myself sanity from rebase crap after master got rebased

* Add safety filter to Kontext.

* Make safety = 2 and input image is optional.

* Add BFL kontext API nodes.

---------

Co-authored-by: Jedrzej Kosinski <kosinkadink1@gmail.com>
2025-05-29 13:27:40 -04:00
comfyanonymous
f2289a1f59 Delete useless file. (#8327) 2025-05-29 08:29:37 -04:00
Robin Huang
fb83eda287 Revert "Add support for Veo3 API node." (#8322)
This reverts commit 592d056100.
2025-05-29 03:03:11 -04:00
Jedrzej Kosinski
5f0e04e2d7 Temporarily adding nodes_v3_test.py file to comfy_extras for testing/sharing purposes 2025-05-28 21:35:14 -07:00
Jedrzej Kosinski
96c2e3856d Add V3-to-V1 compatibility on early V3 node definition and node_info in server.py 2025-05-28 20:56:25 -07:00
comfyanonymous
5e5e46d40c Not really tested WAN Phantom Support. (#8321) 2025-05-28 23:46:15 -04:00
Yoland Yan
4eba3161cf Refactor Pika API node imports and fix unique_id issue. (#8319)
Added unique_id to hidden parameters and corrected description formatting in PikAdditionsNode.
2025-05-28 23:42:25 -04:00
Robin Huang
592d056100 Add support for Veo3 API node. (#8320) 2025-05-28 23:42:02 -04:00
comfyanonymous
1c1687ab1c Support HiDream SimpleTuner loras. (#8318) 2025-05-28 18:47:15 -04:00
comfyanonymous
e6609dacde ComfyUI version 0.3.38 2025-05-28 02:15:11 -04:00
Christian Byrne
ba37e67964 update frontend patch 1.20.7 (#8312) 2025-05-28 01:42:18 -04:00
Jedrzej Kosinski
880f756dc1 More progress on V3 definition 2025-05-27 15:02:17 -07:00
comfyanonymous
06c661004e Memory estimation code can now take into account conds. (#8307) 2025-05-27 15:09:05 -04:00
comfyanonymous
c9e1821a7b ComfyUI version 0.3.37 2025-05-27 07:07:44 -04:00
Jedrzej Kosinski
4480ed488e Initial prototyping on v3 classes 2025-05-25 19:22:42 -07:00
28 changed files with 1825 additions and 91 deletions

View File

@@ -5,20 +5,20 @@
# Inlined the team members for now.
# Maintainers
*.md @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/tests/ @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/tests-unit/ @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/notebooks/ @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/script_examples/ @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/.github/ @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/requirements.txt @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/pyproject.toml @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
*.md @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/tests/ @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/tests-unit/ @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/notebooks/ @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/script_examples/ @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/.github/ @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/requirements.txt @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
/pyproject.toml @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @Kosinkadink @christian-byrne
# Python web server
/api_server/ @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @christian-byrne
/app/ @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @christian-byrne
/utils/ @yoland68 @robinjhuang @huchenlei @webfiltered @pythongosssss @ltdrdata @christian-byrne
/api_server/ @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @christian-byrne
/app/ @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @christian-byrne
/utils/ @yoland68 @robinjhuang @webfiltered @pythongosssss @ltdrdata @christian-byrne
# Node developers
/comfy_extras/ @yoland68 @robinjhuang @huchenlei @pythongosssss @ltdrdata @Kosinkadink @webfiltered @christian-byrne
/comfy/comfy_types/ @yoland68 @robinjhuang @huchenlei @pythongosssss @ltdrdata @Kosinkadink @webfiltered @christian-byrne
/comfy_extras/ @yoland68 @robinjhuang @pythongosssss @ltdrdata @Kosinkadink @webfiltered @christian-byrne
/comfy/comfy_types/ @yoland68 @robinjhuang @pythongosssss @ltdrdata @Kosinkadink @webfiltered @christian-byrne

View File

@@ -205,6 +205,19 @@ comfyui-workflow-templates is not installed.
""".strip()
)
@classmethod
def embedded_docs_path(cls) -> str:
"""Get the path to embedded documentation"""
try:
import comfyui_embedded_docs
return str(
importlib.resources.files(comfyui_embedded_docs) / "docs"
)
except ImportError:
logging.info("comfyui-embedded-docs package not found")
return None
@classmethod
def parse_version_string(cls, value: str) -> tuple[str, str, str]:
"""

View File

@@ -24,6 +24,10 @@ class CONDRegular:
conds.append(x.cond)
return torch.cat(conds)
def size(self):
return list(self.cond.size())
class CONDNoiseShape(CONDRegular):
def process_cond(self, batch_size, device, area, **kwargs):
data = self.cond
@@ -64,6 +68,7 @@ class CONDCrossAttn(CONDRegular):
out.append(c)
return torch.cat(out)
class CONDConstant(CONDRegular):
def __init__(self, cond):
self.cond = cond
@@ -78,3 +83,48 @@ class CONDConstant(CONDRegular):
def concat(self, others):
return self.cond
def size(self):
return [1]
class CONDList(CONDRegular):
def __init__(self, cond):
self.cond = cond
def process_cond(self, batch_size, device, **kwargs):
out = []
for c in self.cond:
out.append(comfy.utils.repeat_to_batch_size(c, batch_size).to(device))
return self._copy_with(out)
def can_concat(self, other):
if len(self.cond) != len(other.cond):
return False
for i in range(len(self.cond)):
if self.cond[i].shape != other.cond[i].shape:
return False
return True
def concat(self, others):
out = []
for i in range(len(self.cond)):
o = [self.cond[i]]
for x in others:
o.append(x.cond[i])
out.append(torch.cat(o))
return out
def size(self): # hackish implementation to make the mem estimation work
o = 0
c = 1
for c in self.cond:
size = c.size()
o += math.prod(size)
if len(size) > 1:
c = size[1]
return [1, c, o // c]

View File

@@ -80,15 +80,13 @@ class DoubleStreamBlock(nn.Module):
(img_mod1, img_mod2), (txt_mod1, txt_mod2) = vec
# prepare image for attention
img_modulated = self.img_norm1(img)
img_modulated = (1 + img_mod1.scale) * img_modulated + img_mod1.shift
img_modulated = torch.addcmul(img_mod1.shift, 1 + img_mod1.scale, self.img_norm1(img))
img_qkv = self.img_attn.qkv(img_modulated)
img_q, img_k, img_v = img_qkv.view(img_qkv.shape[0], img_qkv.shape[1], 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
img_q, img_k = self.img_attn.norm(img_q, img_k, img_v)
# prepare txt for attention
txt_modulated = self.txt_norm1(txt)
txt_modulated = (1 + txt_mod1.scale) * txt_modulated + txt_mod1.shift
txt_modulated = torch.addcmul(txt_mod1.shift, 1 + txt_mod1.scale, self.txt_norm1(txt))
txt_qkv = self.txt_attn.qkv(txt_modulated)
txt_q, txt_k, txt_v = txt_qkv.view(txt_qkv.shape[0], txt_qkv.shape[1], 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
txt_q, txt_k = self.txt_attn.norm(txt_q, txt_k, txt_v)
@@ -102,12 +100,12 @@ class DoubleStreamBlock(nn.Module):
txt_attn, img_attn = attn[:, : txt.shape[1]], attn[:, txt.shape[1] :]
# calculate the img bloks
img = img + img_mod1.gate * self.img_attn.proj(img_attn)
img = img + img_mod2.gate * self.img_mlp((1 + img_mod2.scale) * self.img_norm2(img) + img_mod2.shift)
img.addcmul_(img_mod1.gate, self.img_attn.proj(img_attn))
img.addcmul_(img_mod2.gate, self.img_mlp(torch.addcmul(img_mod2.shift, 1 + img_mod2.scale, self.img_norm2(img))))
# calculate the txt bloks
txt += txt_mod1.gate * self.txt_attn.proj(txt_attn)
txt += txt_mod2.gate * self.txt_mlp((1 + txt_mod2.scale) * self.txt_norm2(txt) + txt_mod2.shift)
txt.addcmul_(txt_mod1.gate, self.txt_attn.proj(txt_attn))
txt.addcmul_(txt_mod2.gate, self.txt_mlp(torch.addcmul(txt_mod2.shift, 1 + txt_mod2.scale, self.txt_norm2(txt))))
if txt.dtype == torch.float16:
txt = torch.nan_to_num(txt, nan=0.0, posinf=65504, neginf=-65504)
@@ -152,7 +150,7 @@ class SingleStreamBlock(nn.Module):
def forward(self, x: Tensor, pe: Tensor, vec: Tensor, attn_mask=None) -> Tensor:
mod = vec
x_mod = (1 + mod.scale) * self.pre_norm(x) + mod.shift
x_mod = torch.addcmul(mod.shift, 1 + mod.scale, self.pre_norm(x))
qkv, mlp = torch.split(self.linear1(x_mod), [3 * self.hidden_size, self.mlp_hidden_dim], dim=-1)
q, k, v = qkv.view(qkv.shape[0], qkv.shape[1], 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
@@ -162,7 +160,7 @@ class SingleStreamBlock(nn.Module):
attn = attention(q, k, v, pe=pe, mask=attn_mask)
# compute activation in mlp stream, cat again and run second linear layer
output = self.linear2(torch.cat((attn, self.mlp_act(mlp)), 2))
x += mod.gate * output
x.addcmul_(mod.gate, output)
if x.dtype == torch.float16:
x = torch.nan_to_num(x, nan=0.0, posinf=65504, neginf=-65504)
return x
@@ -178,6 +176,6 @@ class LastLayer(nn.Module):
shift, scale = vec
shift = shift.squeeze(1)
scale = scale.squeeze(1)
x = (1 + scale[:, None, :]) * self.norm_final(x) + shift[:, None, :]
x = torch.addcmul(shift[:, None, :], 1 + scale[:, None, :], self.norm_final(x))
x = self.linear(x)
return x

View File

@@ -539,13 +539,20 @@ class WanModel(torch.nn.Module):
x = self.unpatchify(x, grid_sizes)
return x
def forward(self, x, timestep, context, clip_fea=None, transformer_options={}, **kwargs):
def forward(self, x, timestep, context, clip_fea=None, time_dim_concat=None, transformer_options={}, **kwargs):
bs, c, t, h, w = x.shape
x = comfy.ldm.common_dit.pad_to_patch_size(x, self.patch_size)
patch_size = self.patch_size
t_len = ((t + (patch_size[0] // 2)) // patch_size[0])
h_len = ((h + (patch_size[1] // 2)) // patch_size[1])
w_len = ((w + (patch_size[2] // 2)) // patch_size[2])
if time_dim_concat is not None:
time_dim_concat = comfy.ldm.common_dit.pad_to_patch_size(time_dim_concat, self.patch_size)
x = torch.cat([x, time_dim_concat], dim=2)
t_len = ((x.shape[2] + (patch_size[0] // 2)) // patch_size[0])
img_ids = torch.zeros((t_len, h_len, w_len, 3), device=x.device, dtype=x.dtype)
img_ids[:, :, :, 0] = img_ids[:, :, :, 0] + torch.linspace(0, t_len - 1, steps=t_len, device=x.device, dtype=x.dtype).reshape(-1, 1, 1)
img_ids[:, :, :, 1] = img_ids[:, :, :, 1] + torch.linspace(0, h_len - 1, steps=h_len, device=x.device, dtype=x.dtype).reshape(1, -1, 1)

View File

@@ -283,8 +283,9 @@ def model_lora_keys_unet(model, key_map={}):
for k in sdk:
if k.startswith("diffusion_model."):
if k.endswith(".weight"):
key_lora = k[len("diffusion_model."):-len(".weight")].replace(".", "_")
key_map["lycoris_{}".format(key_lora)] = k #SimpleTuner lycoris format
key_lora = k[len("diffusion_model."):-len(".weight")]
key_map["lycoris_{}".format(key_lora.replace(".", "_"))] = k #SimpleTuner lycoris format
key_map["transformer.{}".format(key_lora)] = k #SimpleTuner regular format
if isinstance(model, comfy.model_base.ACEStep):
for k in sdk:

View File

@@ -135,6 +135,7 @@ class BaseModel(torch.nn.Module):
logging.info("model_type {}".format(model_type.name))
logging.debug("adm {}".format(self.adm_channels))
self.memory_usage_factor = model_config.memory_usage_factor
self.memory_usage_factor_conds = ()
def apply_model(self, x, t, c_concat=None, c_crossattn=None, control=None, transformer_options={}, **kwargs):
return comfy.patcher_extension.WrapperExecutor.new_class_executor(
@@ -167,6 +168,11 @@ class BaseModel(torch.nn.Module):
if hasattr(extra, "dtype"):
if extra.dtype != torch.int and extra.dtype != torch.long:
extra = extra.to(dtype)
if isinstance(extra, list):
ex = []
for ext in extra:
ex.append(ext.to(dtype))
extra = ex
extra_conds[o] = extra
t = self.process_timestep(t, x=x, **extra_conds)
@@ -325,19 +331,28 @@ class BaseModel(torch.nn.Module):
def scale_latent_inpaint(self, sigma, noise, latent_image, **kwargs):
return self.model_sampling.noise_scaling(sigma.reshape([sigma.shape[0]] + [1] * (len(noise.shape) - 1)), noise, latent_image)
def memory_required(self, input_shape):
def memory_required(self, input_shape, cond_shapes={}):
input_shapes = [input_shape]
for c in self.memory_usage_factor_conds:
shape = cond_shapes.get(c, None)
if shape is not None and len(shape) > 0:
input_shapes += shape
if comfy.model_management.xformers_enabled() or comfy.model_management.pytorch_attention_flash_attention():
dtype = self.get_dtype()
if self.manual_cast_dtype is not None:
dtype = self.manual_cast_dtype
#TODO: this needs to be tweaked
area = input_shape[0] * math.prod(input_shape[2:])
area = sum(map(lambda input_shape: input_shape[0] * math.prod(input_shape[2:]), input_shapes))
return (area * comfy.model_management.dtype_size(dtype) * 0.01 * self.memory_usage_factor) * (1024 * 1024)
else:
#TODO: this formula might be too aggressive since I tweaked the sub-quad and split algorithms to use less memory.
area = input_shape[0] * math.prod(input_shape[2:])
area = sum(map(lambda input_shape: input_shape[0] * math.prod(input_shape[2:]), input_shapes))
return (area * 0.15 * self.memory_usage_factor) * (1024 * 1024)
def extra_conds_shapes(self, **kwargs):
return {}
def unclip_adm(unclip_conditioning, device, noise_augmentor, noise_augment_merge=0.0, seed=None):
adm_inputs = []
@@ -1047,6 +1062,11 @@ class WAN21(BaseModel):
clip_vision_output = kwargs.get("clip_vision_output", None)
if clip_vision_output is not None:
out['clip_fea'] = comfy.conds.CONDRegular(clip_vision_output.penultimate_hidden_states)
time_dim_concat = kwargs.get("time_dim_concat", None)
if time_dim_concat is not None:
out['time_dim_concat'] = comfy.conds.CONDRegular(self.process_latent_in(time_dim_concat))
return out

View File

@@ -297,8 +297,13 @@ except:
try:
if is_amd():
try:
rocm_version = tuple(map(int, str(torch.version.hip).split(".")[:2]))
except:
rocm_version = (6, -1)
arch = torch.cuda.get_device_properties(get_torch_device()).gcnArchName
logging.info("AMD arch: {}".format(arch))
logging.info("ROCm version: {}".format(rocm_version))
if args.use_split_cross_attention == False and args.use_quad_cross_attention == False:
if torch_version_numeric[0] >= 2 and torch_version_numeric[1] >= 7: # works on 2.6 but doesn't actually seem to improve much
if any((a in arch) for a in ["gfx1100", "gfx1101", "gfx1151"]): # TODO: more arches

View File

@@ -1,5 +1,7 @@
from __future__ import annotations
import uuid
import math
import collections
import comfy.model_management
import comfy.conds
import comfy.utils
@@ -104,6 +106,21 @@ def cleanup_additional_models(models):
if hasattr(m, 'cleanup'):
m.cleanup()
def estimate_memory(model, noise_shape, conds):
cond_shapes = collections.defaultdict(list)
cond_shapes_min = {}
for _, cs in conds.items():
for cond in cs:
for k, v in model.model.extra_conds_shapes(**cond).items():
cond_shapes[k].append(v)
if cond_shapes_min.get(k, None) is None:
cond_shapes_min[k] = [v]
elif math.prod(v) > math.prod(cond_shapes_min[k][0]):
cond_shapes_min[k] = [v]
memory_required = model.model.memory_required([noise_shape[0] * 2] + list(noise_shape[1:]), cond_shapes=cond_shapes)
minimum_memory_required = model.model.memory_required([noise_shape[0]] + list(noise_shape[1:]), cond_shapes=cond_shapes_min)
return memory_required, minimum_memory_required
def prepare_sampling(model: ModelPatcher, noise_shape, conds, model_options=None):
executor = comfy.patcher_extension.WrapperExecutor.new_executor(
@@ -117,9 +134,8 @@ def _prepare_sampling(model: ModelPatcher, noise_shape, conds, model_options=Non
models, inference_memory = get_additional_models(conds, model.model_dtype())
models += get_additional_models_from_model_options(model_options)
models += model.get_nested_additional_models() # TODO: does this require inference_memory update?
memory_required = model.memory_required([noise_shape[0] * 2] + list(noise_shape[1:])) + inference_memory
minimum_memory_required = model.memory_required([noise_shape[0]] + list(noise_shape[1:])) + inference_memory
comfy.model_management.load_models_gpu([model] + models, memory_required=memory_required, minimum_memory_required=minimum_memory_required)
memory_required, minimum_memory_required = estimate_memory(model, noise_shape, conds)
comfy.model_management.load_models_gpu([model] + models, memory_required=memory_required + inference_memory, minimum_memory_required=minimum_memory_required + inference_memory)
real_model = model.model
return real_model, conds, models

View File

@@ -256,7 +256,13 @@ def _calc_cond_batch(model: 'BaseModel', conds: list[list[dict]], x_in: torch.Te
for i in range(1, len(to_batch_temp) + 1):
batch_amount = to_batch_temp[:len(to_batch_temp)//i]
input_shape = [len(batch_amount) * first_shape[0]] + list(first_shape)[1:]
if model.memory_required(input_shape) * 1.5 < free_memory:
cond_shapes = collections.defaultdict(list)
for tt in batch_amount:
cond = {k: v.size() for k, v in to_run[tt][0].conditioning.items()}
for k, v in to_run[tt][0].conditioning.items():
cond_shapes[k].append(v.size())
if model.memory_required(input_shape, cond_shapes=cond_shapes) * 1.5 < free_memory:
to_batch = batch_amount
break

View File

@@ -1,25 +0,0 @@
{
"_name_or_path": "openai/clip-vit-large-patch14",
"architectures": [
"CLIPTextModel"
],
"attention_dropout": 0.0,
"bos_token_id": 0,
"dropout": 0.0,
"eos_token_id": 49407,
"hidden_act": "quick_gelu",
"hidden_size": 768,
"initializer_factor": 1.0,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-05,
"max_position_embeddings": 248,
"model_type": "clip_text_model",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"pad_token_id": 1,
"projection_dim": 768,
"torch_dtype": "float32",
"transformers_version": "4.24.0",
"vocab_size": 49408
}

855
comfy_api/v3/io.py Normal file
View File

@@ -0,0 +1,855 @@
from __future__ import annotations
from typing import Any, Literal
from enum import Enum
from abc import ABC, abstractmethod
from dataclasses import dataclass, asdict
from comfy.comfy_types.node_typing import IO
class InputBehavior(str, Enum):
required = "required"
optional = "optional"
def is_class(obj):
'''
Returns True if is a class type.
Returns False if is a class instance.
'''
return isinstance(obj, type)
class NumberDisplay(str, Enum):
number = "number"
slider = "slider"
class IO_V3:
'''
Base class for V3 Inputs and Outputs.
'''
def __init__(self):
pass
def __init_subclass__(cls, io_type: IO | str, **kwargs):
cls.io_type = io_type
super().__init_subclass__(**kwargs)
class InputV3(IO_V3, io_type=None):
'''
Base class for a V3 Input.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None):
super().__init__()
self.id = id
self.display_name = display_name
self.behavior = behavior
self.tooltip = tooltip
self.lazy = lazy
def as_dict_V1(self):
return prune_dict({
"display_name": self.display_name,
"tooltip": self.tooltip,
"lazy": self.lazy
})
def get_io_type_V1(self):
return self.io_type
class WidgetInputV3(InputV3, io_type=None):
'''
Base class for a V3 Input with widget.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None,
default: Any=None,
socketless: bool=None, widgetType: str=None):
super().__init__(id, display_name, behavior, tooltip, lazy)
self.default = default
self.socketless = socketless
self.widgetType = widgetType
def as_dict_V1(self):
return super().as_dict_V1() | prune_dict({
"default": self.default,
"socketless": self.socketless,
"widgetType": self.widgetType,
})
def CustomType(io_type: IO | str) -> type[IO_V3]:
name = f"{io_type}_IO_V3"
return type(name, (IO_V3,), {}, io_type=io_type)
def CustomInput(id: str, io_type: IO | str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None) -> InputV3:
'''
Defines input for 'io_type'. Can be used to stand in for non-core types.
'''
input_kwargs = {
"id": id,
"display_name": display_name,
"behavior": behavior,
"tooltip": tooltip,
"lazy": lazy,
}
return type(f"{io_type}Input", (InputV3,), {}, io_type=io_type)(**input_kwargs)
def CustomOutput(id: str, io_type: IO | str, display_name: str=None, tooltip: str=None) -> OutputV3:
'''
Defines output for 'io_type'. Can be used to stand in for non-core types.
'''
input_kwargs = {
"id": id,
"display_name": display_name,
"tooltip": tooltip,
}
return type(f"{io_type}Output", (OutputV3,), {}, io_type=io_type)(**input_kwargs)
class BooleanInput(WidgetInputV3, io_type=IO.BOOLEAN):
'''
Boolean input.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None,
default: bool=None, label_on: str=None, label_off: str=None,
socketless: bool=None, widgetType: str=None):
super().__init__(id, display_name, behavior, tooltip, lazy, default, socketless, widgetType)
self.label_on = label_on
self.label_off = label_off
self.default: bool
def as_dict_V1(self):
return super().as_dict_V1() | prune_dict({
"label_on": self.label_on,
"label_off": self.label_off,
})
class IntegerInput(WidgetInputV3, io_type=IO.INT):
'''
Integer input.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None,
default: int=None, min: int=None, max: int=None, step: int=None, control_after_generate: bool=None,
display_mode: NumberDisplay=None, socketless: bool=None, widgetType: str=None):
super().__init__(id, display_name, behavior, tooltip, lazy, default, socketless, widgetType)
self.min = min
self.max = max
self.step = step
self.control_after_generate = control_after_generate
self.display_mode = display_mode
self.default: int
def as_dict_V1(self):
return super().as_dict_V1() | prune_dict({
"min": self.min,
"max": self.max,
"step": self.step,
"control_after_generate": self.control_after_generate,
"display": self.display_mode, # NOTE: in frontend, the parameter is called "display"
})
class FloatInput(WidgetInputV3, io_type=IO.FLOAT):
'''
Float input.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None,
default: float=None, min: float=None, max: float=None, step: float=None, round: float=None,
display_mode: NumberDisplay=None, socketless: bool=None, widgetType: str=None):
super().__init__(id, display_name, behavior, tooltip, lazy, default, socketless, widgetType)
self.default = default
self.min = min
self.max = max
self.step = step
self.round = round
self.display_mode = display_mode
self.default: float
def as_dict_V1(self):
return super().as_dict_V1() | prune_dict({
"min": self.min,
"max": self.max,
"step": self.step,
"round": self.round,
"display": self.display_mode, # NOTE: in frontend, the parameter is called "display"
})
class StringInput(WidgetInputV3, io_type=IO.STRING):
'''
String input.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None,
multiline=False, placeholder: str=None, default: int=None,
socketless: bool=None, widgetType: str=None):
super().__init__(id, display_name, behavior, tooltip, lazy, default, socketless, widgetType)
self.multiline = multiline
self.placeholder = placeholder
self.default: str
def as_dict_V1(self):
return super().as_dict_V1() | prune_dict({
"multiline": self.multiline,
"placeholder": self.placeholder,
})
class ComboInput(WidgetInputV3, io_type=IO.COMBO):
'''Combo input (dropdown).'''
def __init__(self, id: str, options: list[str], display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None,
default: str=None, control_after_generate: bool=None,
socketless: bool=None, widgetType: str=None):
super().__init__(id, display_name, behavior, tooltip, lazy, default, socketless, widgetType)
self.multiselect = False
self.options = options
self.control_after_generate = control_after_generate
self.default: str
def as_dict_V1(self):
return super().as_dict_V1() | prune_dict({
"multiselect": self.multiselect,
"options": self.options,
"control_after_generate": self.control_after_generate,
})
class MultiselectComboWidget(ComboInput, io_type=IO.COMBO):
'''Multiselect Combo input (dropdown for selecting potentially more than one value).'''
def __init__(self, id: str, options: list[str], display_name: str=None, behavior=InputBehavior.required, tooltip: str=None, lazy: bool=None,
default: list[str]=None, placeholder: str=None, chip: bool=None, control_after_generate: bool=None,
socketless: bool=None, widgetType: str=None):
super().__init__(id, options, display_name, behavior, tooltip, lazy, default, control_after_generate, socketless, widgetType)
self.multiselect = True
self.placeholder = placeholder
self.chip = chip
self.default: list[str]
def as_dict_V1(self):
return super().as_dict_V1() | prune_dict({
"multiselect": self.multiselect,
"placeholder": self.placeholder,
"chip": self.chip,
})
class ImageInput(InputV3, io_type=IO.IMAGE):
'''
Image input.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None):
super().__init__(id, display_name, behavior, tooltip)
class MaskInput(InputV3, io_type=IO.MASK):
'''
Mask input.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None):
super().__init__(id, display_name, behavior, tooltip)
class LatentInput(InputV3, io_type=IO.LATENT):
'''
Latent input.
'''
def __init__(self, id: str, display_name: str=None, behavior=InputBehavior.required, tooltip: str=None):
super().__init__(id, display_name, behavior, tooltip)
class MultitypedInput(InputV3, io_type="COMFY_MULTITYPED_V3"):
'''
Input that permits more than one input type.
'''
def __init__(self, id: str, io_types: list[type[IO_V3] | InputV3 | IO |str], display_name: str=None, behavior=InputBehavior.required, tooltip: str=None,):
super().__init__(id, display_name, behavior, tooltip)
self._io_types = io_types
@property
def io_types(self) -> list[type[InputV3]]:
'''
Returns list of InputV3 class types permitted.
'''
io_types = []
for x in self._io_types:
if not is_class(x):
io_types.append(type(x))
else:
io_types.append(x)
return io_types
def get_io_type_V1(self):
return ",".join(x.io_type for x in self.io_types)
class OutputV3:
def __init__(self, id: str, display_name: str=None, tooltip: str=None,
is_output_list=False):
self.id = id
self.display_name = display_name
self.tooltip = tooltip
self.is_output_list = is_output_list
def __init_subclass__(cls, io_type, **kwargs):
cls.io_type = io_type
super().__init_subclass__(**kwargs)
class IntegerOutput(OutputV3, io_type=IO.INT):
pass
class FloatOutput(OutputV3, io_type=IO.FLOAT):
pass
class StringOutput(OutputV3, io_type=IO.STRING):
pass
# def __init__(self, id: str, display_name: str=None, tooltip: str=None):
# super().__init__(id, display_name, tooltip)
class ImageOutput(OutputV3, io_type=IO.IMAGE):
pass
class MaskOutput(OutputV3, io_type=IO.MASK):
pass
class LatentOutput(OutputV3, io_type=IO.LATENT):
pass
class DynamicInput(InputV3, io_type=None):
'''
Abstract class for dynamic input registration.
'''
def __init__(self, io_type: str, id: str, display_name: str=None):
super().__init__(io_type, id, display_name)
class DynamicOutput(OutputV3, io_type=None):
'''
Abstract class for dynamic output registration.
'''
def __init__(self, io_type: str, id: str, display_name: str=None):
super().__init__(io_type, id, display_name)
class AutoGrowDynamicInput(DynamicInput, io_type="COMFY_MULTIGROW_V3"):
'''
Dynamic Input that adds another template_input each time one is provided.
Additional inputs are forced to have 'InputBehavior.optional'.
'''
def __init__(self, id: str, template_input: InputV3, min: int=1, max: int=None):
super().__init__("AutoGrowDynamicInput", id)
self.template_input = template_input
if min is not None:
assert(min >= 1)
if max is not None:
assert(max >= 1)
self.min = min
self.max = max
class ComboDynamicInput(DynamicInput, io_type="COMFY_COMBODYNAMIC_V3"):
def __init__(self, id: str):
pass
AutoGrowDynamicInput(id="dynamic", template_input=ImageInput(id="image"))
class Hidden(str, Enum):
'''
Enumerator for requesting hidden variables in nodes.
'''
unique_id = "UNIQUE_ID"
"""UNIQUE_ID is the unique identifier of the node, and matches the id property of the node on the client side. It is commonly used in client-server communications (see messages)."""
prompt = "PROMPT"
"""PROMPT is the complete prompt sent by the client to the server. See the prompt object for a full description."""
extra_pnginfo = "EXTRA_PNGINFO"
"""EXTRA_PNGINFO is a dictionary that will be copied into the metadata of any .png files saved. Custom nodes can store additional information in this dictionary for saving (or as a way to communicate with a downstream node)."""
dynprompt = "DYNPROMPT"
"""DYNPROMPT is an instance of comfy_execution.graph.DynamicPrompt. It differs from PROMPT in that it may mutate during the course of execution in response to Node Expansion."""
auth_token_comfy_org = "AUTH_TOKEN_COMFY_ORG"
"""AUTH_TOKEN_COMFY_ORG is a token acquired from signing into a ComfyOrg account on frontend."""
api_key_comfy_org = "API_KEY_COMFY_ORG"
"""API_KEY_COMFY_ORG is an API Key generated by ComfyOrg that allows skipping signing into a ComfyOrg account on frontend."""
@dataclass
class NodeInfoV1:
input: dict=None
input_order: dict[str, list[str]]=None
output: list[str]=None
output_is_list: list[bool]=None
output_name: list[str]=None
output_tooltips: list[str]=None
name: str=None
display_name: str=None
description: str=None
python_module: Any=None
category: str=None
output_node: bool=None
deprecated: bool=None
experimental: bool=None
api_node: bool=None
def as_pruned_dict(dataclass_obj):
'''Return dict of dataclass object with pruned None values.'''
return prune_dict(asdict(dataclass_obj))
def prune_dict(d: dict):
return {k: v for k,v in d.items() if v is not None}
@dataclass
class SchemaV3:
"""Definition of V3 node properties."""
node_id: str
"""ID of node - should be globally unique. If this is a custom node, add a prefix or postfix to avoid name clashes."""
display_name: str = None
"""Display name of node."""
category: str = "sd"
"""The category of the node, as per the "Add Node" menu."""
inputs: list[InputV3]=None
outputs: list[OutputV3]=None
hidden: list[Hidden]=None
description: str=""
"""Node description, shown as a tooltip when hovering over the node."""
is_input_list: bool = False
"""A flag indicating if this node implements the additional code necessary to deal with OUTPUT_IS_LIST nodes.
All inputs of ``type`` will become ``list[type]``, regardless of how many items are passed in. This also affects ``check_lazy_status``.
From the docs:
A node can also override the default input behaviour and receive the whole list in a single call. This is done by setting a class attribute `INPUT_IS_LIST` to ``True``.
Comfy Docs: https://docs.comfy.org/custom-nodes/backend/lists#list-processing
"""
is_output_node: bool=False
"""Flags this node as an output node, causing any inputs it requires to be executed.
If a node is not connected to any output nodes, that node will not be executed. Usage::
OUTPUT_NODE = True
From the docs:
By default, a node is not considered an output. Set ``OUTPUT_NODE = True`` to specify that it is.
Comfy Docs: https://docs.comfy.org/custom-nodes/backend/server_overview#output-node
"""
is_deprecated: bool=False
"""Flags a node as deprecated, indicating to users that they should find alternatives to this node."""
is_experimental: bool=False
"""Flags a node as experimental, informing users that it may change or not work as expected."""
is_api_node: bool=False
"""Flags a node as an API node. See: https://docs.comfy.org/tutorials/api-nodes/overview."""
# class SchemaV3Class:
# def __init__(self,
# node_id: str,
# node_name: str,
# category: str,
# inputs: list[InputV3],
# outputs: list[OutputV3]=None,
# hidden: list[Hidden]=None,
# description: str="",
# is_input_list: bool = False,
# is_output_node: bool=False,
# is_deprecated: bool=False,
# is_experimental: bool=False,
# is_api_node: bool=False,
# ):
# self.node_id = node_id
# """ID of node - should be globally unique. If this is a custom node, add a prefix or postfix to avoid name clashes."""
# self.node_name = node_name
# """Display name of node."""
# self.category = category
# """The category of the node, as per the "Add Node" menu."""
# self.inputs = inputs
# self.outputs = outputs
# self.hidden = hidden
# self.description = description
# """Node description, shown as a tooltip when hovering over the node."""
# self.is_input_list = is_input_list
# """A flag indicating if this node implements the additional code necessary to deal with OUTPUT_IS_LIST nodes.
# All inputs of ``type`` will become ``list[type]``, regardless of how many items are passed in. This also affects ``check_lazy_status``.
# From the docs:
# A node can also override the default input behaviour and receive the whole list in a single call. This is done by setting a class attribute `INPUT_IS_LIST` to ``True``.
# Comfy Docs: https://docs.comfy.org/custom-nodes/backend/lists#list-processing
# """
# self.is_output_node = is_output_node
# """Flags this node as an output node, causing any inputs it requires to be executed.
# If a node is not connected to any output nodes, that node will not be executed. Usage::
# OUTPUT_NODE = True
# From the docs:
# By default, a node is not considered an output. Set ``OUTPUT_NODE = True`` to specify that it is.
# Comfy Docs: https://docs.comfy.org/custom-nodes/backend/server_overview#output-node
# """
# self.is_deprecated = is_deprecated
# """Flags a node as deprecated, indicating to users that they should find alternatives to this node."""
# self.is_experimental = is_experimental
# """Flags a node as experimental, informing users that it may change or not work as expected."""
# self.is_api_node = is_api_node
# """Flags a node as an API node. See: https://docs.comfy.org/tutorials/api-nodes/overview."""
class classproperty(object):
def __init__(self, f):
self.f = f
def __get__(self, obj, owner):
return self.f(owner)
class ComfyNodeV3(ABC):
"""Common base class for all V3 nodes."""
RELATIVE_PYTHON_MODULE = None
#############################################
# V1 Backwards Compatibility code
#--------------------------------------------
_DESCRIPTION = None
@classproperty
def DESCRIPTION(cls):
if cls._DESCRIPTION is None:
cls.GET_SCHEMA()
return cls._DESCRIPTION
_CATEGORY = None
@classproperty
def CATEGORY(cls):
if cls._CATEGORY is None:
cls.GET_SCHEMA()
return cls._CATEGORY
_EXPERIMENTAL = None
@classproperty
def EXPERIMENTAL(cls):
if cls._EXPERIMENTAL is None:
cls.GET_SCHEMA()
return cls._EXPERIMENTAL
_DEPRECATED = None
@classproperty
def DEPRECATED(cls):
if cls._DEPRECATED is None:
cls.GET_SCHEMA()
return cls._DEPRECATED
_API_NODE = None
@classproperty
def API_NODE(cls):
if cls._API_NODE is None:
cls.GET_SCHEMA()
return cls._API_NODE
_OUTPUT_NODE = None
@classproperty
def OUTPUT_NODE(cls):
if cls._OUTPUT_NODE is None:
cls.GET_SCHEMA()
return cls._OUTPUT_NODE
_INPUT_IS_LIST = None
@classproperty
def INPUT_IS_LIST(cls):
if cls._INPUT_IS_LIST is None:
cls.GET_SCHEMA()
return cls._INPUT_IS_LIST
_OUTPUT_IS_LIST = None
@classproperty
def OUTPUT_IS_LIST(cls):
if cls._OUTPUT_IS_LIST is None:
cls.GET_SCHEMA()
return cls._OUTPUT_IS_LIST
_RETURN_TYPES = None
@classproperty
def RETURN_TYPES(cls):
if cls._RETURN_TYPES is None:
cls.GET_SCHEMA()
return cls._RETURN_TYPES
_RETURN_NAMES = None
@classproperty
def RETURN_NAMES(cls):
if cls._RETURN_NAMES is None:
cls.GET_SCHEMA()
return cls._RETURN_NAMES
_OUTPUT_TOOLTIPS = None
@classproperty
def OUTPUT_TOOLTIPS(cls):
if cls._OUTPUT_TOOLTIPS is None:
cls.GET_SCHEMA()
return cls._OUTPUT_TOOLTIPS
FUNCTION = "execute"
@classmethod
def INPUT_TYPES(cls) -> dict[str, dict]:
schema = cls.DEFINE_SCHEMA()
# for V1, make inputs be a dict with potential keys {required, optional, hidden}
input = {
"required": {}
}
if schema.inputs:
for i in schema.inputs:
input.setdefault(i.behavior.value, {})[i.id] = (i.get_io_type_V1(), i.as_dict_V1())
if schema.hidden:
for hidden in schema.hidden:
input.setdefault("hidden", {})[hidden.name] = (hidden.value,)
return input
@classmethod
def GET_SCHEMA(cls) -> SchemaV3:
schema = cls.DEFINE_SCHEMA()
if cls._DESCRIPTION is None:
cls._DESCRIPTION = schema.description
if cls._CATEGORY is None:
cls._CATEGORY = schema.category
if cls._EXPERIMENTAL is None:
cls._EXPERIMENTAL = schema.is_experimental
if cls._DEPRECATED is None:
cls._DEPRECATED = schema.is_deprecated
if cls._API_NODE is None:
cls._API_NODE = schema.is_api_node
if cls._OUTPUT_NODE is None:
cls._OUTPUT_NODE = schema.is_output_node
if cls._INPUT_IS_LIST is None:
cls._INPUT_IS_LIST = schema.is_input_list
if cls._RETURN_TYPES is None:
output = []
output_name = []
output_is_list = []
output_tooltips = []
if schema.outputs:
for o in schema.outputs:
output.append(o.io_type)
output_name.append(o.display_name if o.display_name else o.io_type)
output_is_list.append(o.is_output_list)
output_tooltips.append(o.tooltip if o.tooltip else None)
cls._RETURN_TYPES = output
cls._RETURN_NAMES = output_name
cls._OUTPUT_IS_LIST = output_is_list
cls._OUTPUT_TOOLTIPS = output_tooltips
return schema
@classmethod
def GET_NODE_INFO_V1(cls) -> dict[str, Any]:
schema = cls.GET_SCHEMA()
# get V1 inputs
input = cls.INPUT_TYPES()
# create separate lists from output fields
output = []
output_is_list = []
output_name = []
output_tooltips = []
if schema.outputs:
for o in schema.outputs:
output.append(o.io_type)
output_is_list.append(o.is_output_list)
output_name.append(o.display_name if o.display_name else o.io_type)
output_tooltips.append(o.tooltip if o.tooltip else None)
info = NodeInfoV1(
input=input,
input_order={key: list(value.keys()) for (key, value) in input.items()},
output=output,
output_is_list=output_is_list,
output_name=output_name,
output_tooltips=output_tooltips,
name=schema.node_id,
display_name=schema.display_name,
category=schema.category,
description=schema.description,
output_node=schema.is_output_node,
deprecated=schema.is_deprecated,
experimental=schema.is_experimental,
api_node=schema.is_api_node,
python_module=getattr(cls, "RELATIVE_PYTHON_MODULE", "nodes")
)
return asdict(info)
#--------------------------------------------
#############################################
@classmethod
def GET_NODE_INFO_V3(cls) -> dict[str, Any]:
schema = cls.GET_SCHEMA()
# TODO: finish
return None
@classmethod
@abstractmethod
def DEFINE_SCHEMA(cls) -> SchemaV3:
"""
Override this function with one that returns a SchemaV3 instance.
"""
return None
DEFINE_SCHEMA = None
def __init__(self):
if self.DEFINE_SCHEMA is None:
raise Exception("No DEFINE_SCHEMA function was defined for this node.")
@abstractmethod
def execute(self, **kwargs) -> NodeOutput:
pass
# class ReturnedInputs:
# def __init__(self):
# pass
# class ReturnedOutputs:
# def __init__(self):
# pass
class NodeOutput:
'''
Standardized output of a node; can pass in any number of args and/or a UIOutput into 'ui' kwarg.
'''
def __init__(self, *args: Any, ui: UIOutput | dict=None, expand: dict=None, block_execution: str=None, **kwargs):
self.args = args
self.ui = ui
self.expand = expand
self.block_execution = block_execution
@property
def result(self):
return self.args if len(self.args) > 0 else None
class SavedResult:
def __init__(self, filename: str, subfolder: str, type: Literal["input", "output", "temp"]):
self.filename = filename
self.subfolder = subfolder
self.type = type
def as_dict(self):
return {
"filename": self.filename,
"subfolder": self.subfolder,
"type": self.type
}
class UIOutput(ABC):
def __init__(self):
pass
@abstractmethod
def as_dict(self) -> dict:
... # TODO: finish
class UIImages(UIOutput):
def __init__(self, values: list[SavedResult | dict], animated=False, **kwargs):
self.values = values
self.animated = animated
def as_dict(self):
values = [x.as_dict() if isinstance(x, SavedResult) else x for x in self.values]
return {
"images": values,
"animated": (self.animated,)
}
class UILatents(UIOutput):
def __init__(self, values: list[SavedResult | dict], **kwargs):
self.values = values
def as_dict(self):
values = [x.as_dict() if isinstance(x, SavedResult) else x for x in self.values]
return {
"latents": values,
}
class UIAudio(UIOutput):
def __init__(self, values: list[SavedResult | dict], **kwargs):
self.values = values
def as_dict(self):
values = [x.as_dict() if isinstance(x, SavedResult) else x for x in self.values]
return {
"audio": values,
}
class UI3D(UIOutput):
def __init__(self, values: list[SavedResult | dict], **kwargs):
self.values = values
def as_dict(self):
values = [x.as_dict() if isinstance(x, SavedResult) else x for x in self.values]
return {
"3d": values,
}
class UIText(UIOutput):
def __init__(self, value: str, **kwargs):
self.value = value
def as_dict(self):
return {"text": (self.value,)}
class TestNode(ComfyNodeV3):
SCHEMA = SchemaV3(
node_id="TestNode_v3",
display_name="Test Node (V3)",
category="v3_test",
inputs=[IntegerInput("my_int"),
#AutoGrowDynamicInput("growing", ImageInput),
MaskInput("thing"),
],
outputs=[ImageOutput("image_output")],
hidden=[Hidden.api_key_comfy_org, Hidden.auth_token_comfy_org, Hidden.unique_id]
)
# @classmethod
# def GET_SCHEMA(cls):
# return cls.SCHEMA
@classmethod
def DEFINE_SCHEMA(cls):
return cls.SCHEMA
def execute(**kwargs):
pass
if __name__ == "__main__":
print("hello there")
inputs: list[InputV3] = [
IntegerInput("my_int"),
CustomInput("xyz", "XYZ"),
CustomInput("model1", "MODEL_M"),
ImageInput("my_image"),
FloatInput("my_float"),
MultitypedInput("my_inputs", [CustomType("MODEL_M"), CustomType("XYZ")]),
]
outputs: list[OutputV3] = [
ImageOutput("image"),
CustomOutput("xyz", "XYZ")
]
for c in inputs:
if isinstance(c, MultitypedInput):
print(f"{c}, {type(c)}, {type(c).io_type}, {c.id}, {[x.io_type for x in c.io_types]}")
print(c.get_io_type_V1())
else:
print(f"{c}, {type(c)}, {type(c).io_type}, {c.id}")
for c in outputs:
print(f"{c}, {type(c)}, {type(c).io_type}, {c.id}")
zz = TestNode()
print(zz.GET_NODE_INFO_V1())
# aa = NodeInfoV1()
# print(asdict(aa))
# print(as_pruned_dict(aa))

View File

@@ -108,6 +108,24 @@ class BFLFluxProGenerateRequest(BaseModel):
# )
class BFLFluxKontextProGenerateRequest(BaseModel):
prompt: str = Field(..., description='The text prompt for what you wannt to edit.')
input_image: Optional[str] = Field(None, description='Image to edit in base64 format')
seed: Optional[int] = Field(None, description='The seed value for reproducibility.')
guidance: confloat(ge=0.1, le=99.0) = Field(..., description='Guidance strength for the image generation process')
steps: conint(ge=1, le=150) = Field(..., description='Number of steps for the image generation process')
safety_tolerance: Optional[conint(ge=0, le=2)] = Field(
2, description='Tolerance level for input and output moderation. Between 0 and 2, 0 being most strict, 6 being least strict. Defaults to 2.'
)
output_format: Optional[BFLOutputFormat] = Field(
BFLOutputFormat.png, description="Output format for the generated image. Can be 'jpeg' or 'png'.", examples=['png']
)
aspect_ratio: Optional[str] = Field(None, description='Aspect ratio of the image between 21:9 and 9:21.')
prompt_upsampling: Optional[bool] = Field(
None, description='Whether to perform upsampling on the prompt. If active, automatically modifies the prompt for more creative generation.'
)
class BFLFluxProUltraGenerateRequest(BaseModel):
prompt: str = Field(..., description='The text prompt for image generation.')
prompt_upsampling: Optional[bool] = Field(

View File

@@ -327,7 +327,9 @@ class ApiClient:
ApiServerError: If the API server is unreachable but internet is working
Exception: For other request failures
"""
url = urljoin(self.base_url, path)
# Use urljoin but ensure path is relative to avoid absolute path behavior
relative_path = path.lstrip('/')
url = urljoin(self.base_url, relative_path)
self.check_auth(self.auth_token, self.comfy_api_key)
# Combine default headers with any provided headers
request_headers = self.get_headers()

View File

@@ -1,6 +1,6 @@
import io
from inspect import cleandoc
from typing import Union
from typing import Union, Optional
from comfy.comfy_types.node_typing import IO, ComfyNodeABC
from comfy_api_nodes.apis.bfl_api import (
BFLStatus,
@@ -9,6 +9,7 @@ from comfy_api_nodes.apis.bfl_api import (
BFLFluxCannyImageRequest,
BFLFluxDepthImageRequest,
BFLFluxProGenerateRequest,
BFLFluxKontextProGenerateRequest,
BFLFluxProUltraGenerateRequest,
BFLFluxProGenerateResponse,
)
@@ -269,6 +270,158 @@ class FluxProUltraImageNode(ComfyNodeABC):
return (output_image,)
class FluxKontextProImageNode(ComfyNodeABC):
"""
Edits images using Flux.1 Kontext [pro] via api based on prompt and aspect ratio.
"""
MINIMUM_RATIO = 1 / 4
MAXIMUM_RATIO = 4 / 1
MINIMUM_RATIO_STR = "1:4"
MAXIMUM_RATIO_STR = "4:1"
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"prompt": (
IO.STRING,
{
"multiline": True,
"default": "",
"tooltip": "Prompt for the image generation - specify what and how to edit.",
},
),
"aspect_ratio": (
IO.STRING,
{
"default": "16:9",
"tooltip": "Aspect ratio of image; must be between 1:4 and 4:1.",
},
),
"guidance": (
IO.FLOAT,
{
"default": 3.0,
"min": 0.1,
"max": 99.0,
"step": 0.1,
"tooltip": "Guidance strength for the image generation process"
},
),
"steps": (
IO.INT,
{
"default": 50,
"min": 1,
"max": 150,
"tooltip": "Number of steps for the image generation process"
},
),
"seed": (
IO.INT,
{
"default": 1234,
"min": 0,
"max": 0xFFFFFFFFFFFFFFFF,
"control_after_generate": True,
"tooltip": "The random seed used for creating the noise.",
},
),
"prompt_upsampling": (
IO.BOOLEAN,
{
"default": False,
"tooltip": "Whether to perform upsampling on the prompt. If active, automatically modifies the prompt for more creative generation, but results are nondeterministic (same seed will not produce exactly the same result).",
},
),
},
"optional": {
"input_image": (IO.IMAGE,),
},
"hidden": {
"auth_token": "AUTH_TOKEN_COMFY_ORG",
"comfy_api_key": "API_KEY_COMFY_ORG",
"unique_id": "UNIQUE_ID",
},
}
@classmethod
def VALIDATE_INPUTS(cls, aspect_ratio: str):
try:
validate_aspect_ratio(
aspect_ratio,
minimum_ratio=cls.MINIMUM_RATIO,
maximum_ratio=cls.MAXIMUM_RATIO,
minimum_ratio_str=cls.MINIMUM_RATIO_STR,
maximum_ratio_str=cls.MAXIMUM_RATIO_STR,
)
except Exception as e:
return str(e)
return True
RETURN_TYPES = (IO.IMAGE,)
DESCRIPTION = cleandoc(__doc__ or "") # Handle potential None value
FUNCTION = "api_call"
API_NODE = True
CATEGORY = "api node/image/BFL"
BFL_PATH = "/proxy/bfl/flux-kontext-pro/generate"
def api_call(
self,
prompt: str,
aspect_ratio: str,
guidance: float,
steps: int,
input_image: Optional[torch.Tensor]=None,
seed=0,
prompt_upsampling=False,
unique_id: Union[str, None] = None,
**kwargs,
):
if input_image is None:
validate_string(prompt, strip_whitespace=False)
operation = SynchronousOperation(
endpoint=ApiEndpoint(
path=self.BFL_PATH,
method=HttpMethod.POST,
request_model=BFLFluxKontextProGenerateRequest,
response_model=BFLFluxProGenerateResponse,
),
request=BFLFluxKontextProGenerateRequest(
prompt=prompt,
prompt_upsampling=prompt_upsampling,
guidance=round(guidance, 1),
steps=steps,
seed=seed,
aspect_ratio=validate_aspect_ratio(
aspect_ratio,
minimum_ratio=self.MINIMUM_RATIO,
maximum_ratio=self.MAXIMUM_RATIO,
minimum_ratio_str=self.MINIMUM_RATIO_STR,
maximum_ratio_str=self.MAXIMUM_RATIO_STR,
),
input_image=(
input_image
if input_image is None
else convert_image_to_base64(input_image)
)
),
auth_kwargs=kwargs,
)
output_image = handle_bfl_synchronous_operation(operation, node_id=unique_id)
return (output_image,)
class FluxKontextMaxImageNode(FluxKontextProImageNode):
"""
Edits images using Flux.1 Kontext [max] via api based on prompt and aspect ratio.
"""
DESCRIPTION = cleandoc(__doc__ or "")
BFL_PATH = "/proxy/bfl/flux-kontext-max/generate"
class FluxProImageNode(ComfyNodeABC):
"""
@@ -914,6 +1067,8 @@ class FluxProDepthNode(ComfyNodeABC):
NODE_CLASS_MAPPINGS = {
"FluxProUltraImageNode": FluxProUltraImageNode,
# "FluxProImageNode": FluxProImageNode,
"FluxKontextProImageNode": FluxKontextProImageNode,
"FluxKontextMaxImageNode": FluxKontextMaxImageNode,
"FluxProExpandNode": FluxProExpandNode,
"FluxProFillNode": FluxProFillNode,
"FluxProCannyNode": FluxProCannyNode,
@@ -924,6 +1079,8 @@ NODE_CLASS_MAPPINGS = {
NODE_DISPLAY_NAME_MAPPINGS = {
"FluxProUltraImageNode": "Flux 1.1 [pro] Ultra Image",
# "FluxProImageNode": "Flux 1.1 [pro] Image",
"FluxKontextProImageNode": "Flux.1 Kontext [pro] Image",
"FluxKontextMaxImageNode": "Flux.1 Kontext [max] Image",
"FluxProExpandNode": "Flux.1 Expand Image",
"FluxProFillNode": "Flux.1 Fill Image",
"FluxProCannyNode": "Flux.1 Canny Control Image",

View File

@@ -6,40 +6,42 @@ Pika API docs: https://pika-827374fb.mintlify.app/api-reference
from __future__ import annotations
import io
from typing import Optional, TypeVar
import logging
import torch
from typing import Optional, TypeVar
import numpy as np
import torch
from comfy.comfy_types.node_typing import IO, ComfyNodeABC, InputTypeOptions
from comfy_api.input_impl import VideoFromFile
from comfy_api.input_impl.video_types import VideoCodec, VideoContainer, VideoInput
from comfy_api_nodes.apinode_utils import (
download_url_to_video_output,
tensor_to_bytesio,
)
from comfy_api_nodes.apis import (
PikaBodyGenerate22T2vGenerate22T2vPost,
PikaGenerateResponse,
PikaBodyGenerate22I2vGenerate22I2vPost,
PikaVideoResponse,
PikaBodyGenerate22C2vGenerate22PikascenesPost,
IngredientsMode,
PikaDurationEnum,
PikaResolutionEnum,
PikaBodyGeneratePikaffectsGeneratePikaffectsPost,
PikaBodyGeneratePikadditionsGeneratePikadditionsPost,
PikaBodyGeneratePikaswapsGeneratePikaswapsPost,
PikaBodyGenerate22C2vGenerate22PikascenesPost,
PikaBodyGenerate22I2vGenerate22I2vPost,
PikaBodyGenerate22KeyframeGenerate22PikaframesPost,
PikaBodyGenerate22T2vGenerate22T2vPost,
PikaBodyGeneratePikadditionsGeneratePikadditionsPost,
PikaBodyGeneratePikaffectsGeneratePikaffectsPost,
PikaBodyGeneratePikaswapsGeneratePikaswapsPost,
PikaDurationEnum,
Pikaffect,
PikaGenerateResponse,
PikaResolutionEnum,
PikaVideoResponse,
)
from comfy_api_nodes.apis.client import (
ApiEndpoint,
HttpMethod,
SynchronousOperation,
PollingOperation,
EmptyRequest,
)
from comfy_api_nodes.apinode_utils import (
tensor_to_bytesio,
download_url_to_video_output,
HttpMethod,
PollingOperation,
SynchronousOperation,
)
from comfy_api_nodes.mapper_utils import model_field_to_node_input
from comfy_api.input_impl.video_types import VideoInput, VideoContainer, VideoCodec
from comfy_api.input_impl import VideoFromFile
from comfy.comfy_types.node_typing import IO, ComfyNodeABC, InputTypeOptions
R = TypeVar("R")
@@ -204,6 +206,7 @@ class PikaImageToVideoV2_2(PikaNodeBase):
"hidden": {
"auth_token": "AUTH_TOKEN_COMFY_ORG",
"comfy_api_key": "API_KEY_COMFY_ORG",
"unique_id": "UNIQUE_ID",
},
}
@@ -457,7 +460,7 @@ class PikAdditionsNode(PikaNodeBase):
},
}
DESCRIPTION = "Add any object or image into your video. Upload a video and specify what youd like to add to create a seamlessly integrated result."
DESCRIPTION = "Add any object or image into your video. Upload a video and specify what you'd like to add to create a seamlessly integrated result."
def api_call(
self,

View File

@@ -14,6 +14,7 @@ import re
from io import BytesIO
from inspect import cleandoc
import torch
import comfy.utils
from comfy.comfy_types import FileLocator
@@ -229,6 +230,186 @@ class SVG:
all_svgs_list.extend(svg_item.data)
return SVG(all_svgs_list)
class ImageStitch:
"""Upstreamed from https://github.com/kijai/ComfyUI-KJNodes"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"image1": ("IMAGE",),
"direction": (["right", "down", "left", "up"], {"default": "right"}),
"match_image_size": ("BOOLEAN", {"default": True}),
"spacing_width": (
"INT",
{"default": 0, "min": 0, "max": 1024, "step": 2},
),
"spacing_color": (
["white", "black", "red", "green", "blue"],
{"default": "white"},
),
},
"optional": {
"image2": ("IMAGE",),
},
}
RETURN_TYPES = ("IMAGE",)
FUNCTION = "stitch"
CATEGORY = "image/transform"
DESCRIPTION = """
Stitches image2 to image1 in the specified direction.
If image2 is not provided, returns image1 unchanged.
Optional spacing can be added between images.
"""
def stitch(
self,
image1,
direction,
match_image_size,
spacing_width,
spacing_color,
image2=None,
):
if image2 is None:
return (image1,)
# Handle batch size differences
if image1.shape[0] != image2.shape[0]:
max_batch = max(image1.shape[0], image2.shape[0])
if image1.shape[0] < max_batch:
image1 = torch.cat(
[image1, image1[-1:].repeat(max_batch - image1.shape[0], 1, 1, 1)]
)
if image2.shape[0] < max_batch:
image2 = torch.cat(
[image2, image2[-1:].repeat(max_batch - image2.shape[0], 1, 1, 1)]
)
# Match image sizes if requested
if match_image_size:
h1, w1 = image1.shape[1:3]
h2, w2 = image2.shape[1:3]
aspect_ratio = w2 / h2
if direction in ["left", "right"]:
target_h, target_w = h1, int(h1 * aspect_ratio)
else: # up, down
target_w, target_h = w1, int(w1 / aspect_ratio)
image2 = comfy.utils.common_upscale(
image2.movedim(-1, 1), target_w, target_h, "lanczos", "disabled"
).movedim(1, -1)
# When not matching sizes, pad to align non-concat dimensions
if not match_image_size:
h1, w1 = image1.shape[1:3]
h2, w2 = image2.shape[1:3]
if direction in ["left", "right"]:
# For horizontal concat, pad heights to match
if h1 != h2:
target_h = max(h1, h2)
if h1 < target_h:
pad_h = target_h - h1
pad_top, pad_bottom = pad_h // 2, pad_h - pad_h // 2
image1 = torch.nn.functional.pad(image1, (0, 0, 0, 0, pad_top, pad_bottom), mode='constant', value=0.0)
if h2 < target_h:
pad_h = target_h - h2
pad_top, pad_bottom = pad_h // 2, pad_h - pad_h // 2
image2 = torch.nn.functional.pad(image2, (0, 0, 0, 0, pad_top, pad_bottom), mode='constant', value=0.0)
else: # up, down
# For vertical concat, pad widths to match
if w1 != w2:
target_w = max(w1, w2)
if w1 < target_w:
pad_w = target_w - w1
pad_left, pad_right = pad_w // 2, pad_w - pad_w // 2
image1 = torch.nn.functional.pad(image1, (0, 0, pad_left, pad_right), mode='constant', value=0.0)
if w2 < target_w:
pad_w = target_w - w2
pad_left, pad_right = pad_w // 2, pad_w - pad_w // 2
image2 = torch.nn.functional.pad(image2, (0, 0, pad_left, pad_right), mode='constant', value=0.0)
# Ensure same number of channels
if image1.shape[-1] != image2.shape[-1]:
max_channels = max(image1.shape[-1], image2.shape[-1])
if image1.shape[-1] < max_channels:
image1 = torch.cat(
[
image1,
torch.ones(
*image1.shape[:-1],
max_channels - image1.shape[-1],
device=image1.device,
),
],
dim=-1,
)
if image2.shape[-1] < max_channels:
image2 = torch.cat(
[
image2,
torch.ones(
*image2.shape[:-1],
max_channels - image2.shape[-1],
device=image2.device,
),
],
dim=-1,
)
# Add spacing if specified
if spacing_width > 0:
spacing_width = spacing_width + (spacing_width % 2) # Ensure even
color_map = {
"white": 1.0,
"black": 0.0,
"red": (1.0, 0.0, 0.0),
"green": (0.0, 1.0, 0.0),
"blue": (0.0, 0.0, 1.0),
}
color_val = color_map[spacing_color]
if direction in ["left", "right"]:
spacing_shape = (
image1.shape[0],
max(image1.shape[1], image2.shape[1]),
spacing_width,
image1.shape[-1],
)
else:
spacing_shape = (
image1.shape[0],
spacing_width,
max(image1.shape[2], image2.shape[2]),
image1.shape[-1],
)
spacing = torch.full(spacing_shape, 0.0, device=image1.device)
if isinstance(color_val, tuple):
for i, c in enumerate(color_val):
if i < spacing.shape[-1]:
spacing[..., i] = c
if spacing.shape[-1] == 4: # Add alpha
spacing[..., 3] = 1.0
else:
spacing[..., : min(3, spacing.shape[-1])] = color_val
if spacing.shape[-1] == 4:
spacing[..., 3] = 1.0
# Concatenate images
images = [image2, image1] if direction in ["left", "up"] else [image1, image2]
if spacing_width > 0:
images.insert(1, spacing)
concat_dim = 2 if direction in ["left", "right"] else 1
return (torch.cat(images, dim=concat_dim),)
class SaveSVGNode:
"""
Save SVG files on disk.
@@ -318,4 +499,5 @@ NODE_CLASS_MAPPINGS = {
"SaveAnimatedWEBP": SaveAnimatedWEBP,
"SaveAnimatedPNG": SaveAnimatedPNG,
"SaveSVGNode": SaveSVGNode,
"ImageStitch": ImageStitch,
}

View File

@@ -296,6 +296,41 @@ class RegexExtract():
return result,
class RegexReplace():
DESCRIPTION = "Find and replace text using regex patterns."
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"string": (IO.STRING, {"multiline": True}),
"regex_pattern": (IO.STRING, {"multiline": True}),
"replace": (IO.STRING, {"multiline": True}),
},
"optional": {
"case_insensitive": (IO.BOOLEAN, {"default": True}),
"multiline": (IO.BOOLEAN, {"default": False}),
"dotall": (IO.BOOLEAN, {"default": False, "tooltip": "When enabled, the dot (.) character will match any character including newline characters. When disabled, dots won't match newlines."}),
"count": (IO.INT, {"default": 0, "min": 0, "max": 100, "tooltip": "Maximum number of replacements to make. Set to 0 to replace all occurrences (default). Set to 1 to replace only the first match, 2 for the first two matches, etc."}),
}
}
RETURN_TYPES = (IO.STRING,)
FUNCTION = "execute"
CATEGORY = "utils/string"
def execute(self, string, regex_pattern, replace, case_insensitive=True, multiline=False, dotall=False, count=0, **kwargs):
flags = 0
if case_insensitive:
flags |= re.IGNORECASE
if multiline:
flags |= re.MULTILINE
if dotall:
flags |= re.DOTALL
result = re.sub(regex_pattern, replace, string, count=count, flags=flags)
return result,
NODE_CLASS_MAPPINGS = {
"StringConcatenate": StringConcatenate,
"StringSubstring": StringSubstring,
@@ -306,7 +341,8 @@ NODE_CLASS_MAPPINGS = {
"StringContains": StringContains,
"StringCompare": StringCompare,
"RegexMatch": RegexMatch,
"RegexExtract": RegexExtract
"RegexExtract": RegexExtract,
"RegexReplace": RegexReplace,
}
NODE_DISPLAY_NAME_MAPPINGS = {
@@ -319,5 +355,6 @@ NODE_DISPLAY_NAME_MAPPINGS = {
"StringContains": "Contains",
"StringCompare": "Compare",
"RegexMatch": "Regex Match",
"RegexExtract": "Regex Extract"
"RegexExtract": "Regex Extract",
"RegexReplace": "Regex Replace",
}

View File

@@ -0,0 +1,67 @@
import torch
from comfy_api.v3.io import (
ComfyNodeV3, SchemaV3, CustomType, CustomInput, CustomOutput, InputBehavior, NumberDisplay,
IntegerInput, MaskInput, ImageInput, ComboDynamicInput, NodeOutput,
)
class V3TestNode(ComfyNodeV3):
@classmethod
def DEFINE_SCHEMA(cls):
return SchemaV3(
node_id="V3TestNode1",
display_name="V3 Test Node (1djekjd)",
description="This is a funky V3 node test.",
category="v3 nodes",
inputs=[
IntegerInput("some_int", display_name="new_name", min=0, tooltip="My tooltip 😎", display_mode=NumberDisplay.slider),
MaskInput("mask", behavior=InputBehavior.optional),
ImageInput("image", display_name="new_image"),
# IntegerInput("some_int", display_name="new_name", min=0, tooltip="My tooltip 😎", display=NumberDisplay.slider, ),
# ComboDynamicInput("mask", behavior=InputBehavior.optional),
# IntegerInput("some_int", display_name="new_name", min=0, tooltip="My tooltip 😎", display=NumberDisplay.slider,
# dependent_inputs=[ComboDynamicInput("mask", behavior=InputBehavior.optional)],
# dependent_values=[lambda my_value: IO.STRING if my_value < 5 else IO.NUMBER],
# ),
# ["option1", "option2". "option3"]
# ComboDynamicInput["sdfgjhl", [ComboDynamicOptions("option1", [IntegerInput("some_int", display_name="new_name", min=0, tooltip="My tooltip 😎", display=NumberDisplay.slider, ImageInput(), MaskInput(), String()]),
# CombyDynamicOptons("option2", [])
# ]]
],
is_output_node=True,
)
def execute(self, some_int: int, image: torch.Tensor, mask: torch.Tensor=None, **kwargs):
a = NodeOutput(1)
aa = NodeOutput(1, "hellothere")
ab = NodeOutput(1, "hellothere", ui={"lol": "jk"})
b = NodeOutput()
c = NodeOutput(ui={"lol": "jk"})
return NodeOutput()
return NodeOutput(1)
return NodeOutput(1, block_execution="Kill yourself")
return ()
NODES_LIST: list[ComfyNodeV3] = [
V3TestNode,
]
# NODE_CLASS_MAPPINGS = {}
# NODE_DISPLAY_NAME_MAPPINGS = {}
# for node in NODES_LIST:
# schema = node.GET_SCHEMA()
# NODE_CLASS_MAPPINGS[schema.node_id] = node
# if schema.display_name:
# NODE_DISPLAY_NAME_MAPPINGS[schema.node_id] = schema.display_name

View File

@@ -345,6 +345,44 @@ class WanCameraImageToVideo:
out_latent["samples"] = latent
return (positive, negative, out_latent)
class WanPhantomSubjectToVideo:
@classmethod
def INPUT_TYPES(s):
return {"required": {"positive": ("CONDITIONING", ),
"negative": ("CONDITIONING", ),
"vae": ("VAE", ),
"width": ("INT", {"default": 832, "min": 16, "max": nodes.MAX_RESOLUTION, "step": 16}),
"height": ("INT", {"default": 480, "min": 16, "max": nodes.MAX_RESOLUTION, "step": 16}),
"length": ("INT", {"default": 81, "min": 1, "max": nodes.MAX_RESOLUTION, "step": 4}),
"batch_size": ("INT", {"default": 1, "min": 1, "max": 4096}),
},
"optional": {"images": ("IMAGE", ),
}}
RETURN_TYPES = ("CONDITIONING", "CONDITIONING", "CONDITIONING", "LATENT")
RETURN_NAMES = ("positive", "negative_text", "negative_img_text", "latent")
FUNCTION = "encode"
CATEGORY = "conditioning/video_models"
def encode(self, positive, negative, vae, width, height, length, batch_size, images):
latent = torch.zeros([batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8], device=comfy.model_management.intermediate_device())
cond2 = negative
if images is not None:
images = comfy.utils.common_upscale(images[:length].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
latent_images = []
for i in images:
latent_images += [vae.encode(i.unsqueeze(0)[:, :, :, :3])]
concat_latent_image = torch.cat(latent_images, dim=2)
positive = node_helpers.conditioning_set_values(positive, {"time_dim_concat": concat_latent_image})
cond2 = node_helpers.conditioning_set_values(negative, {"time_dim_concat": concat_latent_image})
negative = node_helpers.conditioning_set_values(negative, {"time_dim_concat": comfy.latent_formats.Wan21().process_out(torch.zeros_like(concat_latent_image))})
out_latent = {}
out_latent["samples"] = latent
return (positive, cond2, negative, out_latent)
NODE_CLASS_MAPPINGS = {
"WanImageToVideo": WanImageToVideo,
"WanFunControlToVideo": WanFunControlToVideo,
@@ -353,4 +391,5 @@ NODE_CLASS_MAPPINGS = {
"WanVaceToVideo": WanVaceToVideo,
"TrimVideoLatent": TrimVideoLatent,
"WanCameraImageToVideo": WanCameraImageToVideo,
"WanPhantomSubjectToVideo": WanPhantomSubjectToVideo,
}

View File

@@ -1,3 +1,3 @@
# This file is automatically generated by the build process when version is
# updated in pyproject.toml.
__version__ = "0.3.36"
__version__ = "0.3.39"

View File

@@ -17,6 +17,7 @@ from comfy_execution.graph import get_input_info, ExecutionList, DynamicPrompt,
from comfy_execution.graph_utils import is_link, GraphBuilder
from comfy_execution.caching import HierarchicalCache, LRUCache, DependencyAwareCache, CacheKeySetInputSignature, CacheKeySetID
from comfy_execution.validation import validate_node_input
from comfy_api.v3.io import NodeOutput
class ExecutionResult(Enum):
SUCCESS = 0
@@ -242,6 +243,22 @@ def get_output_data(obj, input_data_all, execution_block_cb=None, pre_execute_cb
result = tuple([result] * len(obj.RETURN_TYPES))
results.append(result)
subgraph_results.append((None, result))
elif isinstance(r, NodeOutput):
if r.ui is not None:
uis.append(r.ui.as_dict())
if r.expand is not None:
has_subgraph = True
new_graph = r.expand
result = r.result
if r.block_execution is not None:
result = tuple([ExecutionBlocker(r.block_execution)] * len(obj.RETURN_TYPES))
subgraph_results.append((new_graph, result))
elif r.result is not None:
result = r.result
if r.block_execution is not None:
result = tuple([ExecutionBlocker(r.block_execution)] * len(obj.RETURN_TYPES))
results.append(result)
subgraph_results.append((None, result))
else:
if isinstance(r, ExecutionBlocker):
r = tuple([r] * len(obj.RETURN_TYPES))

View File

@@ -26,6 +26,7 @@ import comfy.sd
import comfy.utils
import comfy.controlnet
from comfy.comfy_types import IO, ComfyNodeABC, InputTypeDict, FileLocator
from comfy_api.v3.io import ComfyNodeV3
import comfy.clip_vision
@@ -2061,6 +2062,7 @@ NODE_DISPLAY_NAME_MAPPINGS = {
"ImagePadForOutpaint": "Pad Image for Outpainting",
"ImageBatch": "Batch Images",
"ImageCrop": "Image Crop",
"ImageStitch": "Image Stitch",
"ImageBlend": "Image Blend",
"ImageBlur": "Image Blur",
"ImageQuantize": "Image Quantize",
@@ -2128,6 +2130,7 @@ def load_custom_node(module_path: str, ignore=set(), module_parent="custom_nodes
if os.path.isdir(web_dir):
EXTENSION_WEB_DIRS[module_name] = web_dir
# V1 node definition
if hasattr(module, "NODE_CLASS_MAPPINGS") and getattr(module, "NODE_CLASS_MAPPINGS") is not None:
for name, node_cls in module.NODE_CLASS_MAPPINGS.items():
if name not in ignore:
@@ -2136,8 +2139,19 @@ def load_custom_node(module_path: str, ignore=set(), module_parent="custom_nodes
if hasattr(module, "NODE_DISPLAY_NAME_MAPPINGS") and getattr(module, "NODE_DISPLAY_NAME_MAPPINGS") is not None:
NODE_DISPLAY_NAME_MAPPINGS.update(module.NODE_DISPLAY_NAME_MAPPINGS)
return True
# V3 node definition
elif getattr(module, "NODES_LIST", None) is not None:
for node_cls in module.NODES_LIST:
node_cls: ComfyNodeV3
schema = node_cls.GET_SCHEMA()
if schema.node_id not in ignore:
NODE_CLASS_MAPPINGS[schema.node_id] = node_cls
node_cls.RELATIVE_PYTHON_MODULE = "{}.{}".format(module_parent, get_module_name(module_path))
if schema.display_name is not None:
NODE_DISPLAY_NAME_MAPPINGS[schema.node_id] = schema.display_name
return True
else:
logging.warning(f"Skip {module_path} module for custom nodes due to the lack of NODE_CLASS_MAPPINGS.")
logging.warning(f"Skip {module_path} module for custom nodes due to the lack of NODE_CLASS_MAPPINGS or NODES_LIST (need one).")
return False
except Exception as e:
logging.warning(traceback.format_exc())
@@ -2257,6 +2271,7 @@ def init_builtin_extra_nodes():
"nodes_ace.py",
"nodes_string.py",
"nodes_camera_trajectory.py",
"nodes_v3_test.py",
]
import_failed = []

View File

@@ -1,6 +1,6 @@
[project]
name = "ComfyUI"
version = "0.3.36"
version = "0.3.39"
readme = "README.md"
license = { file = "LICENSE" }
requires-python = ">=3.9"

View File

@@ -1,5 +1,6 @@
comfyui-frontend-package==1.20.6
comfyui-workflow-templates==0.1.20
comfyui-frontend-package==1.21.3
comfyui-workflow-templates==0.1.23
comfyui-embedded-docs==0.2.0
torch
torchsde
torchvision

View File

@@ -29,6 +29,7 @@ import comfy.model_management
import node_helpers
from comfyui_version import __version__
from app.frontend_management import FrontendManager
from comfy_api.v3.io import ComfyNodeV3
from app.user_manager import UserManager
from app.model_manager import ModelFileManager
@@ -555,6 +556,8 @@ class PromptServer():
def node_info(node_class):
obj_class = nodes.NODE_CLASS_MAPPINGS[node_class]
if isinstance(obj_class, ComfyNodeV3):
return obj_class.GET_NODE_INFO_V1()
info = {}
info['input'] = obj_class.INPUT_TYPES()
info['input_order'] = {key: list(value.keys()) for (key, value) in obj_class.INPUT_TYPES().items()}
@@ -746,6 +749,13 @@ class PromptServer():
web.static('/templates', workflow_templates_path)
])
# Serve embedded documentation from the package
embedded_docs_path = FrontendManager.embedded_docs_path()
if embedded_docs_path:
self.app.add_routes([
web.static('/docs', embedded_docs_path)
])
self.app.add_routes([
web.static('/', self.web_root),
])

View File

View File

@@ -0,0 +1,240 @@
import torch
from unittest.mock import patch, MagicMock
# Mock nodes module to prevent CUDA initialization during import
mock_nodes = MagicMock()
mock_nodes.MAX_RESOLUTION = 16384
with patch.dict('sys.modules', {'nodes': mock_nodes}):
from comfy_extras.nodes_images import ImageStitch
class TestImageStitch:
def create_test_image(self, batch_size=1, height=64, width=64, channels=3):
"""Helper to create test images with specific dimensions"""
return torch.rand(batch_size, height, width, channels)
def test_no_image2_passthrough(self):
"""Test that when image2 is None, image1 is returned unchanged"""
node = ImageStitch()
image1 = self.create_test_image()
result = node.stitch(image1, "right", True, 0, "white", image2=None)
assert len(result) == 1
assert torch.equal(result[0], image1)
def test_basic_horizontal_stitch_right(self):
"""Test basic horizontal stitching to the right"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=32, width=24)
result = node.stitch(image1, "right", False, 0, "white", image2)
assert result[0].shape == (1, 32, 56, 3) # 32 + 24 width
def test_basic_horizontal_stitch_left(self):
"""Test basic horizontal stitching to the left"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=32, width=24)
result = node.stitch(image1, "left", False, 0, "white", image2)
assert result[0].shape == (1, 32, 56, 3) # 24 + 32 width
def test_basic_vertical_stitch_down(self):
"""Test basic vertical stitching downward"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=24, width=32)
result = node.stitch(image1, "down", False, 0, "white", image2)
assert result[0].shape == (1, 56, 32, 3) # 32 + 24 height
def test_basic_vertical_stitch_up(self):
"""Test basic vertical stitching upward"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=24, width=32)
result = node.stitch(image1, "up", False, 0, "white", image2)
assert result[0].shape == (1, 56, 32, 3) # 24 + 32 height
def test_size_matching_horizontal(self):
"""Test size matching for horizontal concatenation"""
node = ImageStitch()
image1 = self.create_test_image(height=64, width=64)
image2 = self.create_test_image(height=32, width=32) # Different aspect ratio
result = node.stitch(image1, "right", True, 0, "white", image2)
# image2 should be resized to match image1's height (64) with preserved aspect ratio
expected_width = 64 + 64 # original + resized (32*64/32 = 64)
assert result[0].shape == (1, 64, expected_width, 3)
def test_size_matching_vertical(self):
"""Test size matching for vertical concatenation"""
node = ImageStitch()
image1 = self.create_test_image(height=64, width=64)
image2 = self.create_test_image(height=32, width=32)
result = node.stitch(image1, "down", True, 0, "white", image2)
# image2 should be resized to match image1's width (64) with preserved aspect ratio
expected_height = 64 + 64 # original + resized (32*64/32 = 64)
assert result[0].shape == (1, expected_height, 64, 3)
def test_padding_for_mismatched_heights_horizontal(self):
"""Test padding when heights don't match in horizontal concatenation"""
node = ImageStitch()
image1 = self.create_test_image(height=64, width=32)
image2 = self.create_test_image(height=48, width=24) # Shorter height
result = node.stitch(image1, "right", False, 0, "white", image2)
# Both images should be padded to height 64
assert result[0].shape == (1, 64, 56, 3) # 32 + 24 width, max(64,48) height
def test_padding_for_mismatched_widths_vertical(self):
"""Test padding when widths don't match in vertical concatenation"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=64)
image2 = self.create_test_image(height=24, width=48) # Narrower width
result = node.stitch(image1, "down", False, 0, "white", image2)
# Both images should be padded to width 64
assert result[0].shape == (1, 56, 64, 3) # 32 + 24 height, max(64,48) width
def test_spacing_horizontal(self):
"""Test spacing addition in horizontal concatenation"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=32, width=24)
spacing_width = 16
result = node.stitch(image1, "right", False, spacing_width, "white", image2)
# Expected width: 32 + 16 (spacing) + 24 = 72
assert result[0].shape == (1, 32, 72, 3)
def test_spacing_vertical(self):
"""Test spacing addition in vertical concatenation"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=24, width=32)
spacing_width = 16
result = node.stitch(image1, "down", False, spacing_width, "white", image2)
# Expected height: 32 + 16 (spacing) + 24 = 72
assert result[0].shape == (1, 72, 32, 3)
def test_spacing_color_values(self):
"""Test that spacing colors are applied correctly"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=32, width=32)
# Test white spacing
result_white = node.stitch(image1, "right", False, 16, "white", image2)
# Check that spacing region contains white values (close to 1.0)
spacing_region = result_white[0][:, :, 32:48, :] # Middle 16 pixels
assert torch.all(spacing_region >= 0.9) # Should be close to white
# Test black spacing
result_black = node.stitch(image1, "right", False, 16, "black", image2)
spacing_region = result_black[0][:, :, 32:48, :]
assert torch.all(spacing_region <= 0.1) # Should be close to black
def test_odd_spacing_width_made_even(self):
"""Test that odd spacing widths are made even"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=32, width=32)
# Use odd spacing width
result = node.stitch(image1, "right", False, 15, "white", image2)
# Should be made even (16), so total width = 32 + 16 + 32 = 80
assert result[0].shape == (1, 32, 80, 3)
def test_batch_size_matching(self):
"""Test that different batch sizes are handled correctly"""
node = ImageStitch()
image1 = self.create_test_image(batch_size=2, height=32, width=32)
image2 = self.create_test_image(batch_size=1, height=32, width=32)
result = node.stitch(image1, "right", False, 0, "white", image2)
# Should match larger batch size
assert result[0].shape == (2, 32, 64, 3)
def test_channel_matching_rgb_to_rgba(self):
"""Test that channel differences are handled (RGB + alpha)"""
node = ImageStitch()
image1 = self.create_test_image(channels=3) # RGB
image2 = self.create_test_image(channels=4) # RGBA
result = node.stitch(image1, "right", False, 0, "white", image2)
# Should have 4 channels (RGBA)
assert result[0].shape[-1] == 4
def test_channel_matching_rgba_to_rgb(self):
"""Test that channel differences are handled (RGBA + RGB)"""
node = ImageStitch()
image1 = self.create_test_image(channels=4) # RGBA
image2 = self.create_test_image(channels=3) # RGB
result = node.stitch(image1, "right", False, 0, "white", image2)
# Should have 4 channels (RGBA)
assert result[0].shape[-1] == 4
def test_all_color_options(self):
"""Test all available color options"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=32, width=32)
colors = ["white", "black", "red", "green", "blue"]
for color in colors:
result = node.stitch(image1, "right", False, 16, color, image2)
assert result[0].shape == (1, 32, 80, 3) # Basic shape check
def test_all_directions(self):
"""Test all direction options"""
node = ImageStitch()
image1 = self.create_test_image(height=32, width=32)
image2 = self.create_test_image(height=32, width=32)
directions = ["right", "left", "up", "down"]
for direction in directions:
result = node.stitch(image1, direction, False, 0, "white", image2)
assert result[0].shape == (1, 32, 64, 3) if direction in ["right", "left"] else (1, 64, 32, 3)
def test_batch_size_channel_spacing_integration(self):
"""Test integration of batch matching, channel matching, size matching, and spacings"""
node = ImageStitch()
image1 = self.create_test_image(batch_size=2, height=64, width=48, channels=3)
image2 = self.create_test_image(batch_size=1, height=32, width=32, channels=4)
result = node.stitch(image1, "right", True, 8, "red", image2)
# Should handle: batch matching, size matching, channel matching, spacing
assert result[0].shape[0] == 2 # Batch size matched
assert result[0].shape[-1] == 4 # Channels matched to max
assert result[0].shape[1] == 64 # Height from image1 (size matching)
# Width should be: 48 + 8 (spacing) + resized_image2_width
expected_image2_width = int(64 * (32/32)) # Resized to height 64
expected_total_width = 48 + 8 + expected_image2_width
assert result[0].shape[2] == expected_total_width