nodes/MeteorAndy-comfyui-byokey
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
master
MeteorAndy-comfyui-byokey/nodes_vidu.py
T
MeteorAndyandClaude 6f2aa13575 Align with official API nodes: VIDEO/AUDIO output + real vendor formats 
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>
2026-06-15 20:07:04 +08:00

278 lines
11 KiB
Python
Raw Permalink Blame History

"""Vidu video generation nodes (bring-your-own-key).
Faithful re-implementations of the official ComfyUI Vidu API nodes
(``ViduTextToVideoNode`` / ``ViduImageToVideoNode``) that call the real Vidu
API directly with the user's own key instead of going through the ComfyOrg
proxy.
Ground truth: ``comfy_api_nodes/nodes_vidu.py`` + ``comfy_api_nodes/apis/vidu.py``.
The official nodes hit ``/proxy/vidu/<path>`` which forwards the body verbatim to
the real Vidu endpoints, so the official request/response models and endpoint
paths ARE the real Vidu API:
POST {base_url}/text2video -> TaskCreationResponse
POST {base_url}/img2video -> TaskCreationResponse
GET {base_url}/tasks/{task_id}/creations -> TaskStatusResponse
Auth header: ``Authorization: Token <api_key>``.
Default host: ``https://api.vidu.com/ent/v2`` (overridable for relays).
"""
from .utils import (
download_video_output,
http_request,
poll_until,
run_async,
tensor_to_data_uri,
validate_prompt,
)
# Real Vidu endpoint host (matches the path the ComfyOrg proxy forwards to).
VIDU_API_BASE = "https://api.vidu.com/ent/v2"
# Endpoint paths (relative to base_url), mirroring the official proxy paths
# VIDU_TEXT_TO_VIDEO = "/proxy/vidu/text2video", etc.
VIDU_TEXT_TO_VIDEO = "/text2video"
VIDU_IMAGE_TO_VIDEO = "/img2video"
VIDU_GET_GENERATION_STATUS = "/tasks/%s/creations"
# Official model / option lists (verbatim from the official node schemas + docs).
VIDU_MODELS = ["viduq1"]
VIDU_ASPECT_RATIOS = ["16:9", "9:16", "1:1"]
VIDU_RESOLUTIONS = ["1080p"]
VIDU_MOVEMENT_AMPLITUDES = ["auto", "small", "medium", "large"]
# Official task states. ``poll_op``'s status_extractor reads ``state``; the
# official success path is reaching a state where ``creations`` is populated.
VIDU_DONE_STATES = ["success", "completed", "succeeded"]
VIDU_FAILED_STATES = ["failed", "error"]
def _auth_headers(api_key: str) -> dict:
"""Build the exact auth headers Vidu expects."""
return {
"Authorization": f"Token {api_key}",
"Content-Type": "application/json",
}
def _build_payload(
*,
model: str,
prompt: str,
duration: int,
seed: int,
aspect_ratio: str | None = None,
resolution: str | None = None,
movement_amplitude: str | None = None,
images: list[str] | None = None,
) -> dict:
"""Assemble a TaskCreationRequest body, omitting unset optional fields.
Mirrors ``comfy_api_nodes.apis.vidu.TaskCreationRequest`` field-for-field
(pydantic drops ``None`` optionals on serialization, so we do the same).
"""
payload: dict = {
"model": model,
"prompt": prompt,
"duration": duration,
"seed": seed,
}
if aspect_ratio is not None:
payload["aspect_ratio"] = aspect_ratio
if resolution is not None:
payload["resolution"] = resolution
if movement_amplitude is not None:
payload["movement_amplitude"] = movement_amplitude
if images is not None:
payload["images"] = images
return payload
async def _execute_task(base_url: str, endpoint: str, headers: dict, payload: dict,
poll_interval: float, max_wait: float):
"""Submit a Vidu task and poll until a video result is available.
Returns a ComfyUI VIDEO object built from the first creation's URL, matching
the official ``execute_task`` + ``download_url_to_video_output`` flow.
"""
base = base_url.rstrip("/")
creation = await http_request("POST", f"{base}{endpoint}", headers, payload)
if str(creation.get("state", "")).lower() == "failed":
raise RuntimeError(f"Vidu request failed. Code: {creation.get('code')}")
task_id = creation.get("task_id") or creation.get("id")
if not task_id:
raise RuntimeError(f"Vidu response did not contain a task_id: {creation}")
status_url = f"{base}{VIDU_GET_GENERATION_STATUS % task_id}"
async def poll_fn():
return await http_request("GET", status_url, headers)
final = await poll_until(
poll_fn,
status_of=lambda r: r.get("state"),
done_statuses=VIDU_DONE_STATES,
failed_statuses=VIDU_FAILED_STATES,
interval=poll_interval,
max_wait=max_wait,
on_fail_message="Vidu generation failed",
)
creations = final.get("creations") or []
if not creations or not creations[0].get("url"):
raise RuntimeError(
f"Vidu request does not contain results. "
f"State: {final.get('state')}, Error Code: {final.get('err_code')}"
)
# Result URLs are presigned/public (valid ~1 hour) and need no auth header,
# exactly like the official download_url_to_video_output.
return await download_video_output(creations[0]["url"])
class BYOKey_Vidu_Text2Video:
"""Vidu text-to-video generation. Output: VIDEO."""
@classmethod
def INPUT_TYPES(cls):
return {
"required": {
"api_key": ("STRING", {"default": "", "multiline": False}),
"base_url": ("STRING", {"default": VIDU_API_BASE, "multiline": False}),
"model": (VIDU_MODELS, {"default": VIDU_MODELS[0]}),
"prompt": ("STRING", {
"multiline": True,
"default": "",
"tooltip": "A textual description for video generation",
}),
},
"optional": {
"duration": ("INT", {
"default": 5, "min": 5, "max": 5, "step": 1,
"tooltip": "Duration of the output video in seconds (fixed at 5)",
}),
"seed": ("INT", {
"default": 0, "min": 0, "max": 2147483647, "step": 1,
"control_after_generate": True,
"tooltip": "Seed for video generation (0 for random)",
}),
"aspect_ratio": (VIDU_ASPECT_RATIOS, {
"default": VIDU_ASPECT_RATIOS[0],
"tooltip": "The aspect ratio of the output video",
}),
"resolution": (VIDU_RESOLUTIONS, {
"default": VIDU_RESOLUTIONS[0],
"tooltip": "Supported values may vary by model & duration",
}),
"movement_amplitude": (VIDU_MOVEMENT_AMPLITUDES, {
"default": VIDU_MOVEMENT_AMPLITUDES[0],
"tooltip": "The movement amplitude of objects in the frame",
}),
"poll_interval": ("INT", {"default": 5, "min": 1, "max": 60, "step": 1}),
"max_wait": ("INT", {"default": 600, "min": 30, "max": 3600, "step": 1}),
},
}
RETURN_TYPES = ("VIDEO",)
RETURN_NAMES = ("output",)
FUNCTION = "execute"
CATEGORY = "api/byokey/video"
DESCRIPTION = "Vidu text-to-video generation (bring your own key)."
def execute(self, api_key, base_url, model, prompt, duration=5, seed=0,
aspect_ratio="16:9", resolution="1080p", movement_amplitude="auto",
poll_interval=5, max_wait=600):
# Official node: prompt is required and cannot be empty.
prompt = validate_prompt(prompt, min_length=1)
headers = _auth_headers(api_key)
payload = _build_payload(
model=model,
prompt=prompt,
duration=duration,
seed=seed,
aspect_ratio=aspect_ratio,
resolution=resolution,
movement_amplitude=movement_amplitude,
)
video = run_async(_execute_task(
base_url, VIDU_TEXT_TO_VIDEO, headers, payload, poll_interval, max_wait,
))
return (video,)
class BYOKey_Vidu_Image2Video:
"""Vidu image-to-video generation. Output: VIDEO."""
@classmethod
def INPUT_TYPES(cls):
return {
"required": {
"api_key": ("STRING", {"default": "", "multiline": False}),
"base_url": ("STRING", {"default": VIDU_API_BASE, "multiline": False}),
"model": (VIDU_MODELS, {"default": VIDU_MODELS[0]}),
"image": ("IMAGE", {
"tooltip": "An image to be used as the start frame of the generated video",
}),
},
"optional": {
"prompt": ("STRING", {
"multiline": True,
"default": "",
"tooltip": "A textual description for video generation",
}),
"duration": ("INT", {
"default": 5, "min": 5, "max": 5, "step": 1,
"tooltip": "Duration of the output video in seconds (fixed at 5)",
}),
"seed": ("INT", {
"default": 0, "min": 0, "max": 2147483647, "step": 1,
"control_after_generate": True,
"tooltip": "Seed for video generation (0 for random)",
}),
"resolution": (VIDU_RESOLUTIONS, {
"default": VIDU_RESOLUTIONS[0],
"tooltip": "Supported values may vary by model & duration",
}),
"movement_amplitude": (VIDU_MOVEMENT_AMPLITUDES, {
"default": VIDU_MOVEMENT_AMPLITUDES[0],
"tooltip": "The movement amplitude of objects in the frame",
}),
"poll_interval": ("INT", {"default": 5, "min": 1, "max": 60, "step": 1}),
"max_wait": ("INT", {"default": 600, "min": 30, "max": 3600, "step": 1}),
},
}
RETURN_TYPES = ("VIDEO",)
RETURN_NAMES = ("output",)
FUNCTION = "execute"
CATEGORY = "api/byokey/video"
DESCRIPTION = "Vidu image-to-video generation (bring your own key)."
def execute(self, api_key, base_url, model, image, prompt="", duration=5, seed=0,
resolution="1080p", movement_amplitude="auto",
poll_interval=5, max_wait=600):
# Official node: only one input image is allowed.
if image.shape[0] > 1:
raise ValueError("Only one input image is allowed.")
# prompt is optional for image-to-video (no validate_prompt).
headers = _auth_headers(api_key)
# Vidu's `images` field accepts a base64 data URI or an image URL.
# Direct API (no ComfyOrg upload service) => inline the image as a data URI.
image_data_uri = tensor_to_data_uri(image, "image/png")
payload = _build_payload(
model=model,
prompt=prompt,
duration=duration,
seed=seed,
resolution=resolution,
movement_amplitude=movement_amplitude,
images=[image_data_uri],
)
video = run_async(_execute_task(
base_url, VIDU_IMAGE_TO_VIDEO, headers, payload, poll_interval, max_wait,
))
return (video,)
Reference in New Issue View Git Blame Copy Permalink