HappyHorse 1.1 视频生成

HappyHorse 1.1 is a video model that lives behind an API, and its three modes are simple enough to explain over a coffee: t2v is pure text-to-video, i2v animates one seed frame, and r2v (reference-to-video) takes up to nine reference images so you can keep the same characters or style across shots. The r2v mode is the reason to seek this node out - you build a little cast sheet of 图1/图2 (image 1/image 2) references and the prompt can literally say "图1 walks toward 图2".

Like every node in T8mars/ComfyUI_Seedance, it's an API wrapper: key from api.seedance.nz, no local weights, requests as the only real dependency.

How it works

Pick a model and the node adapts its inputs. t2v wants just a prompt; i2v grows a required first_image; r2v expands to first_image plus reference_image2 through reference_image9. Gaps get compacted - if you only connect images 1, 3 and 5, the node submits them in connected order, so the "图N" numbering in your prompt follows what you actually connected, not the slot numbers. The aspect ratio you pick is forwarded as metadata.ratio and mapped upstream to the model's own aspectRatio. Everything else is the pack's standard submit → poll → download → decode pipeline.

The inputs that matter

• model - happyhorse-1.1-t2v, happyhorse-1.1-i2v, or happyhorse-1.1-r2v.
• prompt - up to 20480 characters. Required for t2v; optional for i2v; for r2v you can reference 图1..图9 by name.
• seconds - 3 to 15. Note: no -1 "let the model decide" here - HappyHorse doesn't support it, unlike some other nodes in the pack.
• resolution - 720p or 1080p only.
• ratio - forwarded to the upstream aspect-ratio mapping.
• first_image - required for i2v; doubles as 图1 for r2v.
• reference_image2..9 - the rest of your cast.

Outputs: 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, grab a key from api.seedance.nz/console, connect a Seedance API Config node.

Common issues

• r2v ignores your 9th image - it supports up to 9 total (first_image + 8 more), so connect within that limit.
• "图2" in the prompt does nothing - in i2v mode only the first frame exists; the 图N syntax is an r2v feature.
• Seconds=-1 rejected - HappyHorse doesn't take -1; pick 3–15.
• Character drift across a multi-image r2v - that's the model's consistency ceiling, not a wiring bug. If references aren't holding, reorder so the most important character is 图1.

For a closed video model this is a cheap, low-friction way to test character consistency before committing to a pricier render. The pack ships happy-horse example workflows (文生/图生/参考生) you can load straight from the examples/ folder.