☁️BizyAir Seedance 1.0 Pro Image To Video

This is the image-to-video half of ByteDance's closed Seedance line, reached through BizyAir's cloud proxy. Hand it a photo and a prompt describing the motion, and it animates it - no local video model, no VRAM budgeting, just a billed API call. Seedream and Seedance are the two API-only products ByteDance keeps behind a paywall while it open-sources the components underneath them (Depth Anything, PuLID, SDXL-Lightning); this node is a straight line into that closed side.

What it does

first_frame_image is required - this node needs a real starting photo, it doesn't generate from nothing. prompt describes the motion and action you want from that frame. model gives you two variants: doubao-seedance-1-0-pro-250528 (the standard Pro model) and doubao-seedance-1-0-pro-fast-251015 (a later, faster release) - both map to the same doubao-seedance-1-0 family in the output label, so pick fast when you want quicker iteration and standard when you want the better result.

resolution (480p/720p/1080p) and ratio (16:9, 4:3, 1:1, 3:4, 9:16, 21:9, or adaptive) jointly set the output dimensions, same pattern as Seedance's other nodes. duration runs 2–12 seconds, fps is locked at 24 despite showing as a dropdown, and camerafixed is a best-effort request to keep the camera still - the tooltip is explicit that the platform can't guarantee it holds.

The optional last_frame_image is the standout feature here: give it both a start and an end frame, and Seedance interpolates the motion between them instead of extrapolating freely from just the first frame. That's meaningfully more controllable than pure I2V when you know exactly where you want the clip to land.

Notably absent compared to the 1.5 Pro I2V node in this same pack: there's no generate_audio option. Synced audio generation is a 1.5-line capability; 1.0 is silent video only.

seed runs -1 (random) up to a fixed max for reproducible runs.

Output

A video, plus the bizyair_model_name passthrough label.

Installing it

ComfyUI Manager: search BizyAir, install, restart. Manual install:

cd ComfyUI/custom_nodes && git clone https://github.com/siliconflow/BizyAir.git

Restart afterward. No model downloads - this node is a thin client to ByteDance's hosted model via BizyAir. You do need a BizyAir account with an API key set before it'll run; the pack prompts for it on first use.

Where people get burned

Forgetting first_frame_image is required, not optional. Unlike the text-to-video nodes in this pack, this one won't run on a prompt alone.

Picking pro-fast and expecting pro quality. The fast variant trades some fidelity for speed and cost - reach for it when you're iterating on a prompt or motion idea, and switch to the standard Pro model for anything you're actually keeping.

last_frame_image is easy to miss. If you're trying to land a clip on a specific end state and only feeding a start frame, you're leaving Seedance's best controllability feature unused.

It's billed per call. BizyAir runs on paid credits since April 2025 - video generation isn't the cheap end of that pricing, so a batch of exploratory runs adds up quickly.