Nodes/DIGIT Nodes/DIGIT Seedance Video (Replicate) [deprecated]
ComfyUI Node

DIGIT Seedance Video (Replicate) [deprecated]

The DIGIT Seedance node you should probably not be adding anymore

By thedepartmentofexternalservices·Created 6 months ago·Updated 26 days ago· 0
DIGIT Seedance Video (Replicate) [deprecated]
  • first_frame
  • last_frame
  • reference_image1
  • reference_image2
  • reference_image3
  • reference_image4
  • reference_image5
  • reference_image6
  • reference_image7
  • reference_image8
  • reference_image9
  • reference_video1
  • reference_video2
  • reference_video3
  • reference_audio1
  • reference_audio2
  • reference_audio3
  • video
  • video_paths
  • status
prompt
resolution720p
aspect_ratio16:9
duration_seconds5
generate_audiotrue
seed0
negative_prompt

The name on the tin says it all: DIGIT Seedance Video (Replicate) [deprecated]. This is the old Seedance 2.0 node from the DIGIT pack, and the author has moved on - it still works, it still generates video, but you're meant to swap it for the unified DIGIT Seedance Video node (DigitDanceVideo) with the provider dropdown set to replicate. Don't build a new workflow around a node that's scheduled for removal in v5.0.0.

What it does is straightforward: it takes a text prompt and calls ByteDance's Seedance 2.0 video model on Replicate, then drops the finished MP4 back into your graph as a ComfyUI VIDEO. It's an API-wrapper node in the classic sense - nothing runs on your GPU, your prompt and frames go to a third-party service, and you pay per render (the external-api-nodes essay in the modidex KB frames this tradeoff well). For a closed model like Seedance there is no local option, so that's the price of admission.

How it works

Under the hood this is a thin alias. DigitReplicateSeedance is a subclass of DigitDanceVideo that hardcodes provider=replicate and model=seedance-2.0, then forwards your inputs to the same unified backend. That's why the widget surface looks a little bare compared to the current node: no provider dropdown, no model dropdown, no batch count. What you get are the essentials:

  • prompt - what you want to see. In reference mode, you cite inputs with @Image1, @Video1, @Audio1.
  • resolution (480p → 4k) - the tooltip is worth reading: 4k output is 10-bit H.265/HEVC, so it's a bigger, higher-quality encode.
  • aspect_ratio - from 21:9 to 9:16, plus adaptive which lets the model pick.
  • duration_seconds (4–15) - or set -1 to let the model choose.
  • generate_audio, seed.

The optional inputs are where the modes live. Connect first_frame and you get image-to-video; add last_frame and it interpolates between the two; connect any of the nine reference_image / three reference_video / three reference_audio sockets and it's reference-to-video (mutually exclusive with first_frame). negative_prompt is also here - the one thing Replicate's provider supports that fal's doesn't.

Outputs: video (first clip, wire it to a preview or save node), video_paths (all clips), and status (provider, route, cost, request IDs).

Installing and auth

Same as the rest of the pack - ComfyUI Manager, search comfyui-digit, install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/thedepartmentofexternalservices/comfyui-digit.git
cd comfyui-digit
pip install -r requirements.txt

This node specifically needs a Replicate token because it's the Replicate provider:

export REPLICATE_API_TOKEN=your_token

Gotchas

Replicate's Seedance endpoint ships ByteDance's stock content filter, which blocks people and likenesses - if your renders keep coming back rejected, that's the filter, not your prompt. And remember the deprecation: the alias stays registered so old saved workflows keep loading, but the migration path is to open the current DIGIT Seedance Video node, set provider to replicate, and carry on. It logs a warning every run to remind you. Same models, same outputs, and a node that isn't about to vanish.

CategoryDIGIT

Inputs (24)

NameTypeDefaultDescription
promptSTRING
resolutionCOMBO720p4k output is 10-bit H.265/HEVC.
aspect_ratioCOMBO16:97 options: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, +1
duration_secondsINT5-1–15Video length in seconds (4-15). Use -1 for model-selected duration.
generate_audioBOOLEANtrue
seedINT00–2147483647
first_frameoptIMAGEImage-to-video mode. Mutually exclusive with reference inputs.
last_frameoptIMAGEOptional end frame for first-to-last interpolation.
negative_promptoptSTRING
reference_image1optIMAGE
reference_image2optIMAGE
reference_image3optIMAGE
reference_image4optIMAGE
reference_image5optIMAGE
reference_image6optIMAGE
reference_image7optIMAGE
reference_image8optIMAGE
reference_image9optIMAGE
reference_video1optVIDEO
reference_video2optVIDEO
reference_video3optVIDEO
reference_audio1optAUDIO
reference_audio2optAUDIO
reference_audio3optAUDIO

Outputs (3)

NameTypeDescription
videoVIDEO
video_pathsVIDEO_PATHS
statusSTRING