并发提交|MiniMax H3 OW 视频生成(3 合 1)
MiniMax H3 OW in a batch — t2v, i2v, and reference-to-video running together
- image1
- api_config
- future
MiniMax H3 OW is the pack's non-Fast lane into the H3 video family - text-to-video, image-to-video, and reference-to-video, at 480p/720p. If you've been following the H3 story (the KB tracks it as MiniMax's big multimodal video release), you know the model is the real deal; the OW API flavor here is the gateway's take on it. This node is the concurrent-submit wrapper around that 3-in-1, and it's one of the leaner ones in the pack - three models, a small field set, no audio slots.
Wrapper mechanics are the pack standard: validate parameters synchronously, push the job to a background video thread pool, return a SEEDANCE_VIDEO_FUTURE. That feeds SeedanceConcurrent_Video_Await, which waits on up to ten video futures and hands back finished VIDEO values in slot order. Run a t2v and an r2v against the same subject in one pass - two of these, one receiver.
The inputs that matter
model-minimax-h3-ow-t2v,minimax-h3-ow-i2v,minimax-h3-ow-r2v.prompt- required for t2v and r2v; optional for i2v (drive it off the frame if you want).seconds- 5, 10, or 15.resolution-480por720p.ratio- eight options from1:1to21:9.image1- required for i2v and r2v, ignored by t2v. There's exactly one image slot on this node; the multi-image path lives on the Fast sibling.
Plus the pack trio: api_config (or env var / config/.env), skip_error (placeholder video on failure instead of killing the batch), and seed as a ComfyUI cache seed - fixed reuses cached output when nothing else changed, which is how you avoid re-billing an unchanged run.
How it runs
Standard wrapper machinery: upload the reference frame if there is one, submit, poll, download, and only a clip that passes the MP4-header check counts as done. Since validation is synchronous, an i2v/r2v with no image1 fails at the node immediately - before it ever reaches the API or burns a slot in your batch.
Installing
# ComfyUI Manager: search "ComfyUI Seedance" → install → restart.
# or:
comfy node install 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
Only dependency is requests; no weights. API key via Seedance API Config, SEEDANCE_API_KEY, or config/.env.
Where people trip
- i2v/r2v without
image1- bounced at validation. There's no second image slot here, so multi-reference work means switching to the Fast/R2V node. - t2v with
image1attached - ignored, not rejected. Harmless, but the model won't use it. - Nothing but placeholders - that's
skip_error+ a failed slot; check the receiver'sstatus_jsonfor which job died and why.
Standard API-wrapper fine print: per-clip metering, your prompt and frame leave the machine, and result links are temporary - save what you keep.
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| model | COMBO | minimax-h3-ow-t2v | MiniMax H3 OW task type: text-to-video, image-to-video, or reference-image-to-video. | MiniMax H3 OW 支持文生视频、图生视频和参考图生视频。 |
| prompt | STRING | Required for T2V and R2V; optional for I2V. | T2V 与 R2V 必填,I2V 可选。 | |
| seconds | COMBO | 5 | MiniMax H3 OW supports 5, 10, or 15 seconds. | 支持 5、10 或 15 秒。 |
| resolution | COMBO | 480p | MiniMax H3 OW supports 480p or 720p. | 支持 480p 或 720p。 |
| ratio | COMBO | 16:9 | Documented MiniMax H3 OW output aspect ratio. | 文档支持的输出画幅。 |
| image1opt | IMAGE | Required for I2V and R2V; ignored by T2V. | I2V 与 R2V 必须连接,T2V 不使用。 | |
| api_configopt | SEEDANCE_CONFIG | Connect Seedance API Config; otherwise SEEDANCE_API_KEY is used. | |
| skip_erroropt | BOOLEAN | false | On failure return a placeholder error video instead of stopping the workflow. | 失败时输出占位错误视频。 |
| seedopt | INT | 00–18446744073709550000 | ComfyUI cache seed. Fixed reuses the cached result while all other inputs stay unchanged; randomize/increment/decrement starts a new execution. This value is not sent to models without documented seed support. | ComfyUI 缓存种子;Fixed 在其他输入不变时复用缓存,随机、递增或递减会触发新任务。未声明支持 seed 的模型不会收到此参数。 |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| future | SEEDANCE_VIDEO_FUTURE | — |