Vidu2 Image-to-Video Generation

Image-to-video is the default way people actually use Vidu, and this is the Vidu 2 version with a proper model menu. You feed it one start frame, optionally a prompt, and it animates the still into a clip. Compared to the single-model Vidu 2 text node, this one gives you a real decision to make: viduq2-pro, viduq2-pro-fast, or viduq2-turbo. They're the same family at different points on the quality/speed/cost curve.

• viduq2-pro - the flagship. Best motion and consistency, slowest, priciest. Use it for the take you're actually going to ship.
• viduq2-pro-fast - pro quality with a speed discount. The default for most work, honestly; the quality gap to full pro is smaller than the price gap.
• viduq2-turbo - the cheap and cheerful one. Great for iterating on prompts and compositions where you know you'll generate five and keep one.

The pattern is worth internalizing: iterate on turbo, lock a good seed, ship on pro.

The inputs that matter

• image - the start frame. The node validates the aspect ratio must sit between 1:4 and 4:1 (so no extreme panoramas), which covers virtually anything real. The image is where your quality comes from - a strong still gives a strong clip, and no prompt fixes a bad one.
• prompt - optional (max 2000 chars). For image-to-video the best prompts describe the motion, not the picture. The still is already the picture; tell Vidu what happens next.
• duration - 1 to 10 seconds. The slider does what it says.
• resolution - 720p or 1080p. Iterate at 720p, deliver at 1080p.
• movement_amplitude - auto / small / medium / large. The quiet hero control. A talking-head scene wants small; an action shot wants large. If every clip comes back too floaty or too locked-down, this is the knob.
• seed - defaults to 1 (a fixed take); set 0 for random.

How it runs

ComfyUI uploads your image to Vidu's image-to-video endpoint via the Comfy proxy, polls the task, and downloads the finished video. Needs a Comfy account with credits and internet; no GPU, no local model. The output is a VIDEO object - wire it into a save node.

The honest version

The price badge on this node is a small spreadsheet: every model has a per-second rate that changes with resolution, and it's the clearest place in the Vidu family to see why people say "vidu is expensive" in community threads. The pro model at 1080p for 10 seconds is a real-money generation; the community consensus is to treat it like film - spend the expensive pass only after the cheap passes have picked the composition.

And the one rule that governs all of Vidu: same content moderation through ComfyUI as on the website. You're here for the workflow integration - clip lands in your graph, ready for the next node - not for looser rules. For character expression work specifically, the Q3 nodes are where Vidu shines; this node is where you decide whether Vidu's look is for you at all.