Hailuo 2.3 视频生成

Hailuo is one of the video models the local crowd wishes it could run: it's closed, so if you want Hailuo 2.3 clips you call it, and this node is the call. Six models sit behind one dropdown - text-to-video standard and pro, image-to-video standard and pro, and two fast image-to-video variants (standard and pro). The split is the useful part: standard/pro for quality-first renders, fast-* for drafts you don't plan to keep. It's the same API-wrapper arrangement as the rest of the T8mars/ComfyUI_Seedance pack - key from api.seedance.nz, no GPU, no weights.

How it works

Pick a model and the node reshapes to match. Text models want just prompt; image models grow a first_image slot that gets submitted as images[0]. Everything then goes through the pack's standard async pipeline: upload, submit, poll, download, decode to a ComfyUI VIDEO with the usual MP4 header check.

The inputs that matter

• model - the six-way switch described above.
• prompt - up to 2000 characters; required for T2V.
• seconds - 6 or 10. 1080p is locked to 6 seconds, so if you pick 10 you're getting 768p whether you like it or not.
• resolution - 768p or 1080p.
• ratio - text-to-video only; pick adaptive, 16:9, 4:3, 1:1, 3:4, 9:16 or 21:9. Image-to-video ignores this and follows the input image's aspect ratio, which the tooltip says plainly - don't fight it.
• first_image - the seed frame for i2v and fast-i2v.

Outputs are the standard video, video_url, task_id, response.

Installing it

ComfyUI Manager, search "ComfyUI Seedance", or:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/ComfyUI_Seedance.git
cd ../..
python -m pip install -r custom_nodes/ComfyUI_Seedance/requirements.txt

Restart ComfyUI, get a key from api.seedance.nz/console, wire a Seedance API Config node into api_config.

Common issues

• i2v rejects your frame - the source image's short edge must be greater than 300px and its aspect ratio between 2:5 and 5:2. That's a real, verified upstream constraint from the pack's testing, not a suggestion. Resize first.
• Resolution doesn't match what you picked - you set 10 seconds with 1080p; 1080p only runs at 6 seconds.
• Ratio doing nothing on i2v - by design. Image-to-video inherits the frame's shape.

Hailuo 2.3 is the older, cheaper end of the Hailuo line - fine for first drafts and character-tests. If you want H3's 2K output and multimodal references, the pack's separate H3 node is the upgrade path, same install.