MiniMax Hailuo 2.3 Fast (Channel Edition)
Hailuo 2.3 Fast, image-to-video, without a MiniMax account
- first_frame_image
- api_config
- video
- request_id
- urls
MiniMax's Hailuo 2.3 Fast, image-to-video, served through SiliconFlow's BizyAir channel. Hailuo is one of the most popular hosted video models out of China - strong motion, natural physics, a style people actually recognize - and the Fast variant trades a little fidelity for speed. The "Channel Edition" framing means your SiliconFlow key is the only credential you need; you never talk to MiniMax directly.
This is the node for turning a single strong still into a clip fast, when speed matters more than the absolute best quality. If you've got a good image from your local pipeline and you want motion on it in a couple of minutes instead of a local video render session, this is the shortest path in the pack.
How it works
Cloud wrapper, same as every BizyTRD node. Feed it a first_frame_image (required) and a prompt, and the request goes to the BizyAir channel where Hailuo animates the frame. The VIDEO result returns to ComfyUI. The constraints live in the tooltips and they're worth reading: the input image must be JPG/JPEG/PNG/WebP, under 20 MB, with short edges over 300 px and an aspect ratio between 2:5 and 5:2. Prompts can run to 2000 characters and - this is a Hailuo thing - respond to "conch mirror instruction writing," the descriptive prompt style from MiniMax's consumer app, full of camera moves and shot descriptions.
resolution supports 768P and 1080p, but with a catch spelled out in the tooltip: only 768P is available at 10 seconds - 1080p clips max out at 6 seconds. duration is a combo (default 6).
Inputs and outputs
first_frame_image(required) - the frame to animate.prompt(required, multiline, up to 2000 chars).resolution(required, default768P) andduration(required, default 6).- Outputs:
video(VIDEO),request_id,urls.
Installing BizyAirPlus
ComfyUI Manager: "BizyAirPlus", or:
cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt
Restart, account at bizyair.ai, API key, BizyAirPlus button ON. Dependencies: bizyair-cloudberry, bizytrd; no model files.
Common issues
The 20 MB / 300 px / 2:5–5:2 constraints on the first frame are the real traps - an over-ratio panorama or a huge PNG gets rejected with a confusing cloud error. Fix the image upstream, don't fight the node. And the resolution/duration coupling (1080p only at 6s) surprises everyone once. Then the pack standard: missing button → dependency environment (pip show bizyair-cloudberry); cloud failures → key, network, console; and it's metered, so keep an eye on 1080p pricing.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| first_frame_image | IMAGE | Support for JPG/JPEG/PNG/WebP, files smaller than 20MB, short edges larger than 300px, aspect ratio between 2:5 and 5:2. | |
| prompt | STRING | Video generation prompt words, maximum 2000 characters; support conch mirror instruction writing. | |
| resolution | COMBO | 768P | Supports 768P, 1080p; only 768P is supported when the duration is 10 seconds. |
| duration | COMBO | 6 | 6 seconds, 10 seconds; 1080P only supports 6 seconds. |
| api_configopt | BIZYAIR_OPENAPI_CONFIG | 单次请求覆盖base_url和api_key | |
| skip_erroropt | BOOLEAN | false | 开启后遇到错误不中断工作流,输出对应类型的错误占位符 |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| request_id | STRING | — |
| urls | STRING | — |