MeteorAndyandClaude
6f2aa13575
Rewrite each node to faithfully mirror the official comfy_api_nodes/
contract (apis/*.py + Comfy-Org/embedded-docs). The ComfyOrg proxy
forwards the body verbatim, so the official request/response models
ARE the real vendor API.
- Video nodes now return VIDEO (InputImpl.VideoFromFile), not file paths
- Audio nodes now return AUDIO ({waveform, sample_rate}), not file paths
- Request/response bodies match official pydantic models field-for-field
- Model ids, options, defaults, ranges match official + embedded docs
- Every node exposes api_key + base_url (overridable for relays/proxies)
- utils.py: add ComfyUI-native helpers — video_from_bytes,
download_video_output, audio_from_bytes, download_audio_output,
poll_until, validate_prompt, tensor_to_data_uri
- Rewrite nodes_minimax.py (submit -> poll -> retrieve -> download VIDEO)
- nodes_extra.py: base_url on all three, poll_until, validate_prompt
- Clean __init__.py (drop duplicate grok import/keys); 37 nodes total
- README: output-type alignment, node list, relay examples
Co-Authored-By: Claude <noreply@anthropic.com>
403 lines
15 KiB
Python
403 lines
15 KiB
Python
"""Wan (Alibaba DashScope / Tongyi Wanxiang) video generation nodes (bring-your-own-key).
|
||
|
||
These nodes mirror the official ComfyUI Wan API nodes (``WanTextToVideoApi`` /
|
||
``WanImageToVideoApi`` in ``comfy_api_nodes/nodes_wan.py``) but call the Alibaba
|
||
DashScope API DIRECTLY instead of going through the ComfyOrg proxy. The official
|
||
nodes hit ``/proxy/wan/api/v1/...`` which the proxy forwards verbatim to DashScope,
|
||
so the request/response models and endpoint paths used here ARE the real vendor API.
|
||
|
||
DashScope video synthesis is an asynchronous flow:
|
||
1. POST ``/api/v1/services/aigc/video-generation/video-synthesis`` with the
|
||
``X-DashScope-Async: enable`` header -> returns ``output.task_id``.
|
||
2. GET ``/api/v1/tasks/{task_id}`` (Authorization only) -> poll until
|
||
``output.task_status`` is ``SUCCEEDED`` (or ``FAILED``/``UNKNOWN``).
|
||
3. On success, download ``output.video_url`` and wrap it as a ComfyUI VIDEO.
|
||
|
||
Output type is VIDEO (never a string path), matching the official nodes.
|
||
"""
|
||
import logging
|
||
|
||
from .utils import (
|
||
download_video_output,
|
||
http_request,
|
||
poll_until,
|
||
run_async,
|
||
tensor_to_data_uri,
|
||
validate_prompt,
|
||
)
|
||
|
||
logger = logging.getLogger("comfyui_byokey")
|
||
|
||
# Real DashScope host. Path segments mirror the official proxy path
|
||
# (``/proxy/wan`` -> ``https://dashscope.aliyuncs.com``). Overridable for relays
|
||
# (e.g. the international endpoint ``https://dashscope-intl.aliyuncs.com``).
|
||
DEFAULT_BASE_URL = "https://dashscope.aliyuncs.com"
|
||
SUBMIT_PATH = "/api/v1/services/aigc/video-generation/video-synthesis"
|
||
TASK_PATH = "/api/v1/tasks/{task_id}"
|
||
|
||
# Resolution / aspect-ratio combos for the size-driven text-to-video node, copied
|
||
# verbatim from the official WanTextToVideoApi schema. The "(WIDTHxHEIGHT)" tail is
|
||
# parsed out and sent to DashScope as the "WIDTH*HEIGHT" size string.
|
||
T2V_SIZE_OPTIONS = [
|
||
"480p: 1:1 (624x624)",
|
||
"480p: 16:9 (832x480)",
|
||
"480p: 9:16 (480x832)",
|
||
"720p: 1:1 (960x960)",
|
||
"720p: 16:9 (1280x720)",
|
||
"720p: 9:16 (720x1280)",
|
||
"720p: 4:3 (1088x832)",
|
||
"720p: 3:4 (832x1088)",
|
||
"1080p: 1:1 (1440x1440)",
|
||
"1080p: 16:9 (1920x1080)",
|
||
"1080p: 9:16 (1080x1920)",
|
||
"1080p: 4:3 (1632x1248)",
|
||
"1080p: 3:4 (1248x1632)",
|
||
]
|
||
|
||
# Resolution labels for the image-to-video node (official WanImageToVideoApi).
|
||
I2V_RESOLUTION_OPTIONS = ["480P", "720P", "1080P"]
|
||
|
||
SHOT_TYPE_OPTIONS = ["single", "multi"]
|
||
|
||
# Task status values returned by DashScope under output.task_status.
|
||
DONE_STATUSES = ["SUCCEEDED"]
|
||
FAILED_STATUSES = ["FAILED", "UNKNOWN", "CANCELED"]
|
||
|
||
|
||
def _parse_size_from_label(size_label: str) -> str:
|
||
"""Extract ``WIDTH*HEIGHT`` from a label like ``"720p: 16:9 (1280x720)"``.
|
||
|
||
DashScope expects the ``size`` parameter as ``"<width>*<height>"``.
|
||
"""
|
||
start = size_label.rfind("(")
|
||
end = size_label.rfind(")")
|
||
if start == -1 or end == -1 or end <= start:
|
||
raise ValueError(f"Could not parse resolution from size label: {size_label!r}")
|
||
inner = size_label[start + 1:end]
|
||
for sep in ("x", "X", "×"): # x, X, ×
|
||
if sep in inner:
|
||
w, h = inner.split(sep, 1)
|
||
return f"{w.strip()}*{h.strip()}"
|
||
raise ValueError(f"Could not parse resolution from size label: {size_label!r}")
|
||
|
||
|
||
def _auth_headers(api_key: str, *, async_submit: bool = False) -> dict:
|
||
"""Build DashScope auth headers. The async header is required on submit only."""
|
||
headers = {
|
||
"Authorization": f"Bearer {api_key}",
|
||
"Content-Type": "application/json",
|
||
}
|
||
if async_submit:
|
||
headers["X-DashScope-Async"] = "enable"
|
||
return headers
|
||
|
||
|
||
async def _submit_and_wait(
|
||
base_url: str,
|
||
api_key: str,
|
||
body: dict,
|
||
poll_interval: int,
|
||
max_wait: int,
|
||
):
|
||
"""Submit a DashScope video task, poll until done, return the VIDEO output."""
|
||
base = base_url.rstrip("/")
|
||
submit_url = f"{base}{SUBMIT_PATH}"
|
||
|
||
submit_resp = await http_request(
|
||
"POST", submit_url, _auth_headers(api_key, async_submit=True), json_data=body
|
||
)
|
||
output = submit_resp.get("output") or {}
|
||
task_id = output.get("task_id")
|
||
if not task_id:
|
||
code = submit_resp.get("code")
|
||
message = submit_resp.get("message")
|
||
raise RuntimeError(
|
||
f"DashScope did not return a task_id. code={code} message={message} response={submit_resp}"
|
||
)
|
||
|
||
task_url = f"{base}{TASK_PATH.format(task_id=task_id)}"
|
||
poll_headers = _auth_headers(api_key)
|
||
|
||
async def _poll():
|
||
return await http_request("GET", task_url, poll_headers)
|
||
|
||
final = await poll_until(
|
||
_poll,
|
||
status_of=lambda r: (r.get("output") or {}).get("task_status"),
|
||
done_statuses=DONE_STATUSES,
|
||
failed_statuses=FAILED_STATUSES,
|
||
interval=poll_interval,
|
||
max_wait=max_wait,
|
||
on_fail_message="Wan video generation failed",
|
||
)
|
||
|
||
final_output = final.get("output") or {}
|
||
video_url = final_output.get("video_url")
|
||
if not video_url:
|
||
raise RuntimeError(f"Wan task succeeded but no video_url was returned: {final}")
|
||
|
||
# The download URL is authenticated on DashScope's CDN and does not need the
|
||
# API key, so download without auth headers.
|
||
return await download_video_output(video_url)
|
||
|
||
|
||
class BYOKey_Wan_Text2Video:
|
||
"""Wan text-to-video via Alibaba DashScope (bring-your-own-key).
|
||
|
||
Mirrors the official ``WanTextToVideoApi`` node. Output: VIDEO.
|
||
"""
|
||
|
||
@classmethod
|
||
def INPUT_TYPES(cls):
|
||
return {
|
||
"required": {
|
||
"api_key": ("STRING", {
|
||
"default": "",
|
||
"tooltip": "DashScope (Alibaba Model Studio) API key.",
|
||
}),
|
||
"model": (["wan2.5-t2v-preview", "wan2.6-t2v"], {
|
||
"default": "wan2.6-t2v",
|
||
"tooltip": "Model to use.",
|
||
}),
|
||
"prompt": ("STRING", {
|
||
"multiline": True,
|
||
"default": "",
|
||
"tooltip": "Prompt describing the elements and visual features. "
|
||
"Supports English and Chinese.",
|
||
}),
|
||
},
|
||
"optional": {
|
||
"base_url": ("STRING", {
|
||
"default": DEFAULT_BASE_URL,
|
||
"tooltip": "DashScope API host. Use https://dashscope-intl.aliyuncs.com "
|
||
"for the international endpoint, or a relay base URL.",
|
||
}),
|
||
"negative_prompt": ("STRING", {
|
||
"multiline": True,
|
||
"default": "",
|
||
"tooltip": "Negative prompt describing what to avoid.",
|
||
}),
|
||
"size": (T2V_SIZE_OPTIONS, {
|
||
"default": "720p: 1:1 (960x960)",
|
||
"tooltip": "Video resolution and aspect ratio.",
|
||
}),
|
||
"duration": ("INT", {
|
||
"default": 5,
|
||
"min": 5,
|
||
"max": 15,
|
||
"step": 5,
|
||
"tooltip": "Video duration in seconds. A 15-second duration is "
|
||
"available only for the Wan 2.6 model.",
|
||
}),
|
||
"seed": ("INT", {"default": 0, "min": 0, "max": 2147483647}),
|
||
"generate_audio": ("BOOLEAN", {
|
||
"default": False,
|
||
"tooltip": "If no audio input is provided, generate audio automatically.",
|
||
}),
|
||
"prompt_extend": ("BOOLEAN", {
|
||
"default": True,
|
||
"tooltip": "Whether to enhance the prompt with AI assistance.",
|
||
}),
|
||
"watermark": ("BOOLEAN", {
|
||
"default": False,
|
||
"tooltip": "Whether to add an AI-generated watermark to the result.",
|
||
}),
|
||
"shot_type": (SHOT_TYPE_OPTIONS, {
|
||
"default": "single",
|
||
"tooltip": "Single continuous shot or multiple shots with cuts. "
|
||
"Takes effect only when prompt_extend is True.",
|
||
}),
|
||
"poll_interval": ("INT", {"default": 6, "min": 2, "max": 60}),
|
||
"max_wait": ("INT", {"default": 600, "min": 60, "max": 3600}),
|
||
},
|
||
}
|
||
|
||
RETURN_TYPES = ("VIDEO",)
|
||
RETURN_NAMES = ("output",)
|
||
FUNCTION = "execute"
|
||
CATEGORY = "api/byokey/video"
|
||
DESCRIPTION = "Wan text-to-video generation via Alibaba DashScope (bring your own key)."
|
||
|
||
def execute(
|
||
self,
|
||
api_key,
|
||
model,
|
||
prompt,
|
||
base_url=DEFAULT_BASE_URL,
|
||
negative_prompt="",
|
||
size="720p: 1:1 (960x960)",
|
||
duration=5,
|
||
seed=0,
|
||
generate_audio=False,
|
||
prompt_extend=True,
|
||
watermark=False,
|
||
shot_type="single",
|
||
poll_interval=6,
|
||
max_wait=600,
|
||
):
|
||
validate_prompt(prompt, min_length=1)
|
||
if "480p" in size and model == "wan2.6-t2v":
|
||
raise ValueError("The Wan 2.6 model does not support 480p.")
|
||
if duration == 15 and model == "wan2.5-t2v-preview":
|
||
raise ValueError("A 15-second duration is supported only by the Wan 2.6 model.")
|
||
|
||
size_value = _parse_size_from_label(size)
|
||
|
||
input_field = {"prompt": prompt}
|
||
if negative_prompt:
|
||
input_field["negative_prompt"] = negative_prompt
|
||
|
||
body = {
|
||
"model": model,
|
||
"input": input_field,
|
||
"parameters": {
|
||
"size": size_value,
|
||
"duration": duration,
|
||
"seed": seed,
|
||
"audio": generate_audio,
|
||
"prompt_extend": prompt_extend,
|
||
"watermark": watermark,
|
||
"shot_type": shot_type,
|
||
},
|
||
}
|
||
|
||
video = run_async(
|
||
_submit_and_wait(base_url, api_key, body, poll_interval, max_wait)
|
||
)
|
||
return (video,)
|
||
|
||
|
||
class BYOKey_Wan_Image2Video:
|
||
"""Wan image-to-video via Alibaba DashScope (bring-your-own-key).
|
||
|
||
Mirrors the official ``WanImageToVideoApi`` node. The input image is the first
|
||
frame; the output video's aspect ratio is derived from it. Output: VIDEO.
|
||
"""
|
||
|
||
@classmethod
|
||
def INPUT_TYPES(cls):
|
||
return {
|
||
"required": {
|
||
"api_key": ("STRING", {
|
||
"default": "",
|
||
"tooltip": "DashScope (Alibaba Model Studio) API key.",
|
||
}),
|
||
"model": (["wan2.5-i2v-preview", "wan2.6-i2v"], {
|
||
"default": "wan2.6-i2v",
|
||
"tooltip": "Model to use.",
|
||
}),
|
||
"image": ("IMAGE", {
|
||
"tooltip": "Input image used as the first frame. Exactly one image is required.",
|
||
}),
|
||
"prompt": ("STRING", {
|
||
"multiline": True,
|
||
"default": "",
|
||
"tooltip": "Prompt describing the elements and visual features. "
|
||
"Supports English and Chinese.",
|
||
}),
|
||
},
|
||
"optional": {
|
||
"base_url": ("STRING", {
|
||
"default": DEFAULT_BASE_URL,
|
||
"tooltip": "DashScope API host. Use https://dashscope-intl.aliyuncs.com "
|
||
"for the international endpoint, or a relay base URL.",
|
||
}),
|
||
"negative_prompt": ("STRING", {
|
||
"multiline": True,
|
||
"default": "",
|
||
"tooltip": "Negative prompt describing what to avoid.",
|
||
}),
|
||
"resolution": (I2V_RESOLUTION_OPTIONS, {
|
||
"default": "720P",
|
||
"tooltip": "Video resolution quality. The Wan 2.6 model does not support 480P.",
|
||
}),
|
||
"duration": ("INT", {
|
||
"default": 5,
|
||
"min": 5,
|
||
"max": 15,
|
||
"step": 5,
|
||
"tooltip": "Video duration in seconds. A 15-second duration is "
|
||
"supported only by the Wan 2.6 model.",
|
||
}),
|
||
"seed": ("INT", {"default": 0, "min": 0, "max": 2147483647}),
|
||
"generate_audio": ("BOOLEAN", {
|
||
"default": False,
|
||
"tooltip": "If no audio input is provided, generate audio automatically.",
|
||
}),
|
||
"prompt_extend": ("BOOLEAN", {
|
||
"default": True,
|
||
"tooltip": "Whether to enhance the prompt with AI assistance.",
|
||
}),
|
||
"watermark": ("BOOLEAN", {
|
||
"default": False,
|
||
"tooltip": "Whether to add an AI-generated watermark to the result.",
|
||
}),
|
||
"shot_type": (SHOT_TYPE_OPTIONS, {
|
||
"default": "single",
|
||
"tooltip": "Single continuous shot or multiple shots with cuts. "
|
||
"Takes effect only when prompt_extend is True.",
|
||
}),
|
||
"poll_interval": ("INT", {"default": 6, "min": 2, "max": 60}),
|
||
"max_wait": ("INT", {"default": 600, "min": 60, "max": 3600}),
|
||
},
|
||
}
|
||
|
||
RETURN_TYPES = ("VIDEO",)
|
||
RETURN_NAMES = ("output",)
|
||
FUNCTION = "execute"
|
||
CATEGORY = "api/byokey/video"
|
||
DESCRIPTION = "Wan image-to-video generation via Alibaba DashScope (bring your own key)."
|
||
|
||
def execute(
|
||
self,
|
||
api_key,
|
||
model,
|
||
image,
|
||
prompt,
|
||
base_url=DEFAULT_BASE_URL,
|
||
negative_prompt="",
|
||
resolution="720P",
|
||
duration=5,
|
||
seed=0,
|
||
generate_audio=False,
|
||
prompt_extend=True,
|
||
watermark=False,
|
||
shot_type="single",
|
||
poll_interval=6,
|
||
max_wait=600,
|
||
):
|
||
validate_prompt(prompt, min_length=1)
|
||
if image.shape[0] != 1:
|
||
raise ValueError(
|
||
f"Exactly one input image is required, but got a batch of {image.shape[0]}."
|
||
)
|
||
if "480P" in resolution and model == "wan2.6-i2v":
|
||
raise ValueError("The Wan 2.6 model does not support 480P.")
|
||
if duration == 15 and model == "wan2.5-i2v-preview":
|
||
raise ValueError("A 15-second duration is supported only by the Wan 2.6 model.")
|
||
|
||
img_url = tensor_to_data_uri(image, mime_type="image/png")
|
||
|
||
input_field = {"prompt": prompt, "img_url": img_url}
|
||
if negative_prompt:
|
||
input_field["negative_prompt"] = negative_prompt
|
||
|
||
body = {
|
||
"model": model,
|
||
"input": input_field,
|
||
"parameters": {
|
||
"resolution": resolution,
|
||
"duration": duration,
|
||
"seed": seed,
|
||
"audio": generate_audio,
|
||
"prompt_extend": prompt_extend,
|
||
"watermark": watermark,
|
||
"shot_type": shot_type,
|
||
},
|
||
}
|
||
|
||
video = run_async(
|
||
_submit_and_wait(base_url, api_key, body, poll_interval, max_wait)
|
||
)
|
||
return (video,)
|