Alexander Garzon
e8fdd92f21
Manual testing in ComfyUI surfaced two defects in the gpt-image-2.5 work.
background=transparent was silently useless: the API returned a correct RGBA
PNG, but pil_to_tensor ran convert("RGB") on every response and composited the
alpha onto black. Both nodes now output (IMAGE, MASK), with the alpha carried on
the mask as 1 - alpha to match the LoadImage convention, so Join Image with
Alpha consumes it directly. The three response decode helpers collapsed into
response_item_to_pil, since the old path already decoded each item to a PIL
image and then discarded the alpha -- routing through a single decode was
smaller than adding a second one.
CUSTOM_DIM_MIN was 1024 while validate_custom_dimensions() accepts a short edge
down to 480 (1440x480 at the 3:1 aspect limit). The widget was a second, stricter
gate that disagreed with the validator, so legal sizes like 1536x864 snapped up
to 1024 and were unreachable from the UI. The floor is now derived from what the
validator accepts.
The existing tests missed both: they assert on request kwargs and never checked
that alpha survives the response, and they called run_generate() directly,
bypassing the widget bounds entirely.
Verified against the live API (transparent output recomposes to a fully
transparent corner pixel) and by manual testing in ComfyUI.
77 lines
4.0 KiB
Python
77 lines
4.0 KiB
Python
"""ComfyUI node: image edit / inpaint via OpenAI images.edit.
|
|
|
|
Like node_generate, `model` is a DynamicCombo with a nested size combo. Reference
|
|
images use an auto-growing input (up to 16). Behavior lives in build.run_edit().
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
from comfy_api.latest import IO
|
|
|
|
from . import build
|
|
from . import params as params_mod
|
|
from .node_generate import _gpt_image_2_inputs, _gpt_image_25_inputs, unpack_size
|
|
|
|
_MAX_REFS = 16
|
|
|
|
|
|
class OpenAIImageEdit(IO.ComfyNode):
|
|
@classmethod
|
|
def define_schema(cls):
|
|
return IO.Schema(
|
|
node_id="ImagentOpenAIImageEdit",
|
|
display_name="🤖 Imagent: OpenAI Image Edit",
|
|
category="Imagent",
|
|
description="BYOK image edit / inpaint / multi-reference compose via OpenAI "
|
|
"gpt-image models (direct API, no proxy).",
|
|
inputs=[
|
|
IO.String.Input("prompt", multiline=True, default="",
|
|
placeholder="Describe the edit to apply…",
|
|
tooltip="Describe the edit to apply to the reference image(s)."),
|
|
IO.DynamicCombo.Input(
|
|
"model",
|
|
tooltip="OpenAI gpt-image model. Switching this changes the options below.",
|
|
options=[
|
|
IO.DynamicCombo.Option("gpt-image-2.5-flare", _gpt_image_25_inputs()),
|
|
IO.DynamicCombo.Option("gpt-image-2.5-sunburst", _gpt_image_25_inputs()),
|
|
IO.DynamicCombo.Option("gpt-image-2", _gpt_image_2_inputs()),
|
|
],
|
|
),
|
|
IO.Autogrow.Input(
|
|
"images",
|
|
template=IO.Autogrow.TemplateNames(
|
|
IO.Image.Input("image"),
|
|
names=[f"image_{i}" for i in range(1, _MAX_REFS + 1)],
|
|
min=1),
|
|
tooltip=f"Reference image(s) to edit; grows up to {_MAX_REFS}. A mask "
|
|
"(if used) applies to the first image and requires exactly one image."),
|
|
IO.Combo.Input("quality", options=params_mod.QUALITIES, default="auto",
|
|
tooltip="Rendering quality. 'auto' lets the model decide; "
|
|
"higher quality costs more. 'xhigh' and 'max' need "
|
|
"gpt-image-2.5 and fall back to 'high' elsewhere."),
|
|
IO.Combo.Input("output_format", options=params_mod.FORMATS, default="png",
|
|
tooltip="Image file format returned by the API."),
|
|
IO.Combo.Input("moderation", options=params_mod.MODERATIONS, default="auto",
|
|
tooltip="Content moderation strictness. 'auto' (default) or "
|
|
"'low'. Sent only when not 'auto'."),
|
|
IO.Int.Input("n", default=1, min=params_mod.N_MIN, max=params_mod.N_MAX, step=1,
|
|
tooltip="How many images to generate (1-8)."),
|
|
IO.Mask.Input("mask", optional=True,
|
|
tooltip="Inpaint mask: white marks the region to edit. Requires a "
|
|
"single reference image."),
|
|
],
|
|
outputs=[IO.Image.Output(display_name="image"),
|
|
IO.Mask.Output(display_name="mask")],
|
|
)
|
|
|
|
@classmethod
|
|
def execute(cls, prompt, model, images, quality, output_format, moderation, n,
|
|
mask=None) -> IO.NodeOutput:
|
|
size, custom_width, custom_height = unpack_size(model)
|
|
ref_tensors = [t for t in (images or {}).values() if t is not None]
|
|
img, out_mask, _info = build.run_edit(
|
|
prompt=prompt, model=model["model"], size=size, quality=quality,
|
|
background=model["background"], output_format=output_format, n=n,
|
|
images=ref_tensors, mask=mask, moderation=moderation,
|
|
custom_width=custom_width, custom_height=custom_height)
|
|
return IO.NodeOutput(img, out_mask)
|