☁️BizyAir Seedance 1.5 Pro Image To Video

The image-to-video node for ByteDance's newer Seedance 1.5 Pro model (doubao-seedance-1-5-pro-251215), reached through BizyAir's cloud API. Same job as the 1.0 I2V node elsewhere in this pack - animate a still photo from a prompt - but 1.5 adds one thing 1.0 doesn't have: synced audio generation. Seedance is one half of ByteDance's closed-product line (Seedream for images, Seedance for video); the components ByteDance actually gives away for free - Depth Anything, PuLID, SDXL-Lightning - live somewhere else entirely.

What it does

first_frame_image is required, so this node needs a real starting photo - it's not a text-to-video tool. prompt describes the motion and, now, the audio. resolution (480p/720p/1080p) and ratio (16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive) jointly determine output dimensions. duration runs 2–12 seconds, fps is fixed at 24 despite the dropdown, and camerafixed is a best-effort "try to keep the camera still" flag with no guarantee attached.

The optional last_frame_image lets you set an end frame alongside the start, so Seedance interpolates between two fixed points instead of freely extrapolating - genuinely the most controllable way to use this node if you know where the clip should land.

generate_audio is the headline addition over the 1.0 line: Seedance 1.5 Pro can generate voice, sound effects, and background music synced to what's happening on screen, driven from your prompt text. The tooltip's own advice is to wrap spoken dialogue in quotes for better results - and it flags explicitly that turning this on affects billing, so it's not a no-cost toggle.

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

Output

A video, plus the bizyair_model_name passthrough label.

Installing it

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

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

Restart after. No local downloads - the generation runs on ByteDance's infrastructure, proxied through BizyAir, so what you actually need is a BizyAir account with an API key set (the pack handles first-time setup with a login link).

Where people get burned

Quality reports on Seedance 1.5 are mixed in practice. Real users on the API have flagged rough motion coherence on anything beyond simple actions - worth a cheap test generation before committing a batch of credits to a complex prompt.

Audio quietly adds cost. If a bill looks higher than expected, generate_audio left on from a previous run is the first thing to check.

last_frame_image is the feature people skip. If you want the clip to land on a specific pose or state rather than drift wherever the model decides, feeding an end frame is the whole point of choosing this node over the plain 1.0 line.

Forgetting first_frame_image is mandatory. This node animates a photo; it doesn't generate video from a prompt alone the way the T2V node does.