BYOKey Vidu Image to Video

Vidu is the Chinese video API that flies under most people's radar, and its image-to-video is about as minimal as this pack gets. You give it one image, optionally a prompt, and it animates the frame into a five-second clip on the viduq1 model. No resolution menus to argue with (it's 1080p, full stop), no duration picker (five seconds, also full stop). The whole design is "feed it a frame and go."

That minimalism is the appeal. When you want a fast, predictable i2v pass over a batch of images - turn a few stills into preview clips - this is the least fiddly door in the pack.

How it works. Same submit-then-poll pattern as every video node here. POST to api.vidu.com/ent/v2 (the default base_url, overridable for a relay) with your Vidu key, with the image sent as an inline base64 data URI and the optional prompt alongside it. Poll every poll_interval seconds until the clip is done, download the mp4, and wrap it in ComfyUI's native VIDEO type.

The inputs that matter:

• image - required. The start frame of the video. Wire in anything with an IMAGE output.
• prompt - optional, but worth writing anyway: a sentence about the motion ("she looks up slowly", "camera drifts right") does a lot more than a scene description, since the scene is already in the frame.
• movement_amplitude - auto, small, medium, or large. This is Vidu's distinctive knob: how much motion happens in the frame. Small for subtle, moody life; large for dramatic movement. auto lets the model decide.
• duration - fixed at 5, with min and max both set to 5. It's not a bug, it's the current model contract; don't go hunting for a longer option.
• seed - 0 for random; Vidu honors it if you set one.
• resolution - 1080p, also the only choice. Supported values may vary by model and duration, the tooltip warns, but today there's one.
• poll_interval / max_wait - the usual babysitters; the 10-minute cap is generous for a 5-second clip.

Output: one output (VIDEO) socket, into your usual video preview/save node.

Install. Shared pack, one clone:

cd ComfyUI/custom_nodes
git clone https://github.com/MeteorAndy/comfyui-byokey.git
# restart ComfyUI; it's under api/byokey/video

No model downloads, no extra dependencies beyond ComfyUI's bundled torch, aiohttp, PIL, numpy and av.

Gotchas. The fixed 5-second duration catches people who expected a range - it's the API, not the node, and it'll change when Vidu's model does. And the pack is young with no community footprint, so the standing rule applies: skim the source before feeding it a key.