seedance
ByteDance's video engine on Civitai's fleet
- images
- api_config
- video
- workflow_id
- raw_json
This is the node that puts ByteDance's Seedance video engine inside your ComfyUI graph - and unlike most of what you run in ComfyUI, none of it happens on your machine. Seedance is closed and API-only (that's ByteDance's whole posture: the products stay on the API, only the components go open). This civitai-comfy-nodes recipe node submits your job to Civitai's Orchestration API, waits, and hands the finished clip back as a native VIDEO socket. Billed in Buzz.
Of all the video nodes in this pack, Seedance is the most feature-complete out of the box. It's the one you reach for when you want real choices - resolution, duration, aspect ratio, audio, even reference inputs - without fighting a dozen fiddly knobs.
The inputs that matter
Three are required, and they're the core of the recipe:
prompt- the usual multiline prompt.aspect_ratio- seven options including16:9,9:16,1:1,4:3,3:4,21:9, andadaptive(which matches your input image's ratio - great for image-to-video).duration- 4 to 15 seconds, the longest range of any node in this pack.resolution-480p,720p, or1080p.
Then the optional stack, where the interesting stuff lives:
model-v2,v2-fast, orv2-minion top of the empty default.v2-fasttrades a little quality for a shorter queue and smaller bill;v2-miniis the cheap draft tier.generate_audio(default on) - Seedance can synthesize audio alongside the video. If you don't want it, flip it off and save the Buzz; if you do, this is one of the few cloud video engines here that handles sound natively.images- wire aLoadImagein for image-to-video. This is the input that does the most for your output quality: anchoring the first frame to a known image fixes the consistency problems that plague pure text-to-video.reference_videos_jsonandreference_audios_json- JSON text fields for character/subject reference. These are the pack's "freeform structures stay JSON" inputs from the README, so you pass an array of URLs. Invalid JSON fails fast with a clear local error telling you exactly what's malformed.seed- set it to lock a result, or leave at 0.api_config- auth.
Outputs
The standard pack trio: video (VIDEO) into a Save Video / preview node, workflow_id (STRING) for tracking in the Civitai sidebar, raw_json (STRING) to inspect the response - including what each run billed in Buzz.
Install and auth
One-time, pack-wide setup. ComfyUI Manager → Custom Nodes Manager → search Civitai Comfy Nodes → install, restart. Or from the shell:
cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txt # just `requests`
Auth: wire a Civitai Auth node, set CIVITAI_API_TOKEN, or sign in via the sidebar. Headless boxes should use the env var.
Common issues
- Cost is per-second and it adds up. The community's honest reaction to Seedance-in-ComfyUI was sticker shock - video jobs are where a cloud session gets expensive fast. Check
raw_jsonafter a run to see the meter before you chain ten of these together. - Slow at launch. The pack shipped June 2026, right as Civitai's datacenter lost GPU capacity to crypto mining; long queues and the occasional timeout were the top real-world complaints. The default timeout is 30 minutes - raise it via the Civitai Auth node if long jobs keep dying.
- Same filter as the web generator. ByteDance's content moderation applies at the model, not the node. There's no local bypass.
- Reference inputs must be valid JSON. If you get a "not valid JSON" error from
reference_videos_json, it's a bracket or quote problem in your text field, not a network failure.
It's not the budget option in this pack - MiniMax or Mochi are cheaper per clip - but Seedance is the closest thing here to an all-rounder: I2V, audio, long durations, and a real resolution ladder.
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | — | |
| aspect_ratio | COMBO | 7 options: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, +1 | |
| duration | COMBO | 12 options: 4, 5, 6, 7, 8, 9, +6 | |
| resolution | COMBO | 3 options: 480p, 720p, 1080p | |
| modelopt | COMBO | 4 options: , v2, v2-fast, v2-mini | |
| generate_audioopt | BOOLEAN | true | — |
| seedopt | INT | 00–4294967295 | — |
| imagesopt | IMAGE | — | |
| reference_videos_jsonopt | STRING | — | |
| reference_audios_jsonopt | STRING | — | |
| api_configopt | CIVITAI_CONFIG | Optional Civitai Auth connection; defaults to CIVITAI_API_TOKEN or stored OAuth login. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| workflow_id | STRING | — |
| raw_json | STRING | — |