Google Veo3.1 Fast Image To Video Fast (Channel Edition)

Image in, prompt in, clip out - the image-to-video side of Google's Veo 3.1 Fast, in its Channel Edition. It's the sibling of the text-to-video version of this node, differing in exactly one way that matters: it has an images input, so the video starts from a frame you already like rather than from nothing.

Veo 3.1 is the closed video model the open-source community keeps measuring itself against, and one of the few closed models with a genuine capability gap - native audio in the output. The Fast tier is the speed-and-cost-focused variant, and this Channel Edition keeps the surface minimal: images, prompt, resolution (720p by default), aspect_ratio (16:9). No duration, no seed, no audio toggle. If you want a quick animation of a still you already generated locally, this is the least ceremony available.

Why you'd reach for it

Because animating a good frame is the classic workflow move, and doing it locally usually means a video model, a sampler config, and a wait that depends on your GPU. This node outsources the whole problem: wire a Load Image into images, describe the motion in prompt, queue. The result drops back as a normal video output you preview or save like any local render. The trade is the standard one for this pack - a billed, closed, cloud model - but for "make my still image move" it's a hard workflow to beat for convenience.

How it works

The images port is a COMFY_AUTOGROW_V3 input, so it accepts the connected image cleanly. The prompt field is your chance to direct the motion - subject, action, environment, camera movement, and audio cues are exactly what the field wants. On queue, the node uploads your image, submits the request through BizyAir to Veo 3.1 Fast, polls, and returns video (VIDEO), request_id (STRING), and urls (STRING). The optional aspect_ratio defaults to 16:9.

Optional robustness inputs: api_config (BIZYAIR_OPENAPI_CONFIG) overrides base_url and api_key per request; skip_error turns a failed call into an error placeholder instead of an aborted workflow.

Install and setup

One-time install, same for the entire pack:

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

Restart ComfyUI (or install "BizyAirPlus" via ComfyUI Manager). No model files - the engine arrives as the bizyair-cloudberry and bizytrd pip packages auto-installed on startup. Click the BizyAirPlus action-bar button to ON, get an API key from bizyair.ai, enter it, and queue.

Common issues

• Motion is weak or weird - image-to-video models amplify whatever the source frame implies. A static, blurry input gives a static, blurry clip; start from a frame with obvious motion cues.
• Nothing returns - the switch is off or the key is dead; the console has the details.
• Billed for every try - each queue is a generation. Get the prompt and frame right before you queue, not after.

Pack caveat as always: BizyAir auto-installs and sends your image to the cloud. It's a legitimate SiliconFlow product, but flip the switch OFF when you're only running local workflows.