Nodes/PopNow/Seedance 2.0 Mini · Video Extend
ComfyUI Node

Seedance 2.0 Mini · Video Extend

More seconds without starting over

By popnow-ai·Created 24 days ago·Updated about 20 hours ago· 0
Seedance 2.0 Mini · Video Extend
  • video_url
  • last_image_url
  • video
  • task id
◄prompt►
◄resolution720p►
◄duration_seconds5►
◄generate_audiotrue►
◄enable_web_searchfalse►
◄seed0►
◄bitrate_modestandard►
◄watermarkfalse►

Continuation, at the cheap end of the ladder. You give it a clip, it renders a new segment starting from the last frame, and PopNow hands back the joined result - your footage plus the new seconds, one file, source not billed.

Where this earns its keep is the case every video workflow eventually hits: the take you like is three seconds too short, and regenerating means losing it. Where it hurts is the case people discover afterwards - extend enough times and the shot drifts, because each new segment starts from a generated frame rather than real footage.

How it works

Upload, quote, submit under an idempotency key, poll, download. MP4 sources are remuxed on the way up rather than re-encoded, so the join starts from the quality you actually made. Your source accepts MP4, MOV or WebM up to 60 seconds, and there's a 90 MiB upload ceiling, which is the practical limit for longer or higher-bitrate files.

The re-basing detail is the one to internalise: segment two's opening frame is segment one's generated last frame. Errors in anatomy, texture, colour and lighting compound from there. Chain twice and you're usually still fine. Chain five times and the shot has quietly become something else.

The inputs that matter

video_url - the clip to continue. From a video loader, or straight off another PopNow node's video output.

prompt - what happens next: the action that follows, the camera move, the sound. The tooltip frames it as a description of the continuation, not a re-description of the shot, and it's required here (unlike Fast's extend, where it's optional). Use it.

duration_seconds - the length of the new segment, 4 to 15. The source clip is added on top and isn't billed, so this number is the entire cost. Mini's version is a combo rather than a slider, but the values are the same seconds.

resolution - 480p or 720p. Match the source; a resolution mismatch at the join is the most visible artifact you can produce with this node.

generate_audio - sound for the new segment only; the source keeps its own. That's the right design and it's why extending a clip with good audio doesn't cost you a re-score.

enable_web_search - web lookup before rendering, billed at the web-search rate. There's rarely a reason for it on a continuation.

bitrate_mode - standard or high for the new segment. watermark - off by default. seed - fixed, so a re-queue returns the same clip instead of charging you twice.

last_image_url - an optional target end frame. The new segment lands on that picture, which is a cheap way to make a continuation actually go somewhere.

Outputs

video - the joined clip. task id - the job id. No image output on this node, so if you want a still from the end of the extension, save the video and pull a frame.

Installing it

Manager → PopNow → Install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/popnow-ai/popnow-comfyui

Key from popnow.ai/api/console/keys into Settings > PopNow (or POPNOW_API_KEY in the environment), then a second restart so the catalog is read from your account. Nothing else to install - this whole pack is a client, not a runtime.

Where people get burned

Drift, as covered: two extensions is a reasonable ceiling for anything you care about, and the fix is generating the length up front rather than extending a fourth time. This isn't specific to Mini or to PopNow - it's what happens when a video model is asked to continue its own output.

The 90 MiB ceiling catches people extending long clips, since a 60-second source can exceed it before you've added anything. Re-encode down first; retrying won't help.

And because extensions are individually cheap, they're individually easy to spam. Set Settings > PopNow > Max credits per run, and remember that a cancelled run keeps rendering on the server and bills if it succeeds.

CategoryPopNow/Video/ByteDance

Inputs (10)

NameTypeDefaultDescription
promptSTRINGDescribe what happens next: the action that follows, the camera move and the sound.
video_urlVIDEOThe clip to continue: MP4, MOV or WebM, up to 60 seconds. Its final frame becomes the opening frame of the new segment, and the result is your clip with the new segment joined on.
resolutionCOMBO720pResolution of the new segment; match your source clip.
duration_secondsCOMBO5Length of the new segment in seconds. The source clip is added on top and is not billed.
generate_audioBOOLEANtrueGenerate sound for the new segment. The source keeps its own sound either way.
enable_web_searchBOOLEANfalseLet the model search the web for information related to the prompt before rendering. Requests with it on are billed at the web-search rate.
seedINT00–2147483647Same seed and settings, same result. Set 'control after generate' to randomize to get a new result on every run; every run is billed.
bitrate_modeCOMBOstandardEncoding bitrate of the new segment: standard, or high for crisper detail and a larger file.
watermarkBOOLEANfalseAdd the model’s visible "AI generated" watermark to the new segment.
last_image_urloptIMAGEOptional target end frame: the new segment lands on this picture.

Outputs (2)

NameTypeDescription
videoVIDEO—
task idSTRINGThe PopNow task id, for the API console logs or support.