ByteDance Seedance 2.0 Mini First Last Frame (Official Edition)

Give this node a first frame, optionally a last frame, and it returns a video that animates from one to the other - ByteDance's Seedance 2.0 Mini running in the cloud through BizyAirPlus, in its Official Edition (the vendor's direct API route rather than BizyAir's channel). It's the same first-last-frame trick the community uses to turn static images they actually like into clips that keep the composition.

The FLF workflow is the cure for text-to-video's main failure mode: you write a great prompt, the model ignores half of it, and you get a video of the wrong thing. With first/last-frame you've pinned the shot. The model's only job is motion between two points, which is a much smaller ask - and on the cheap Mini tier, a much smaller risk.

How it works

Output node in the BizyAirPlus cloud layer. The BizyAirPlus ON button in the action bar puts the workflow in cloud mode; the node uploads your frames plus the prompt and duration to Seedance's service, and the rendered clip comes back into ComfyUI as a normal video. Closed model, so nothing is ever downloaded locally. The Mini designation is Seedance's fast/budget tier - good enough for drafts and short clips, not the place to demand film-quality rendering.

The inputs that matter

• first_frame_url - required. The starting image, typed IMAGE, so wire it straight from a Load Image node or your image-generation output.
• prompt - describes the motion between the frames. The frames decide what, the prompt decides how.
• duration - 4 to 15 seconds (default 8).
• last_frame_url - optional. Skip it and Seedance invents the ending; provide it and the video lands exactly on it.
• resolution - 480p default.
• aspect_ratio, seed, generate_audio - the usual suspects. generate_audio defaults on; seed = -1 is random.

This official Mini node doesn't carry a watermark toggle or web search, so it's the leaner sibling of the Channel Mini FLF node - slightly fewer switches, same core.

The outputs

video goes to a preview or save node. urls is the direct download link. request_id is the ID to keep for support when something fails mid-generation.

Installing it

In BizyAirPlus. Manager install (search "BizyAirPlus") or:

cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt

Restart ComfyUI, turn BizyAirPlus ON, enter the key from bizyair.ai. Saved to ~/.BizyAirPlus/apikey.ini, or set BIZYAIR_API_KEY. The heavy lifting lives in the bizytrd and bizyair-cloudberry pip packages - no model files, no VRAM requirements, nothing to babysit.

Gotchas

First-run errors are almost always the button being OFF or a missing key; the ComfyUI console has the real error message. Don't panic if you've read things calling BizyAir malware - that was a ComfyUI Manager auto-install mixup from older versions, since fixed, and the pack is a legit SiliconFlow product. For the node itself: choose bookend frames that belong to the same shot. Mismatched lighting or subject detail makes the model compromise, and the compromise is visible in the middle frames.