Command Palette
Search for a command to run...
google/veo-3Veo 3.1 Fast — Google DeepMind's fast, cost-efficient Veo 3.1 video model with native audio (dialogue and sound effects), image-to-video, and up to 1080p resolution.
curl https://api.runbase.net/v1/runs \
-H "Authorization: Bearer $RUNBASE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "google/veo-3",
"input": {
"prompt": "A cinematic product photo of a ceramic lamp",
"aspect_ratio": "1:1",
"resolution": "1K"
}
}'
First-person view soaring low over a medieval battlefield at dawn, gliding past clashing knights in armor, fire-lit arrows whizzing overhead, splintered catapults burning near fallen soldiers, flying inches above torn flags and mud-soaked ground, raw, terrifying, epic
Veo 3.1 Fast is the speed- and cost-optimized variant of Google DeepMind's Veo 3.1 video model. Like the rest of the Veo 3 family, it generates audio natively — producing synchronized dialogue, sound effects, and ambient audio alongside the visuals in a single pass. The model generates video at up to 1080p in two aspect ratios (16:9 and 9:16) and accepts up to two frame images for image-to-video generation, enabling first-frame and last-frame control.
Cinematic content where audio is integral — nature documentaries with ambient sound, product reveals with impact effects. Social reels and vertical video with native sound design. Image-to-video with start and end frames for controlled animation arcs. Dialogue scenes and character-driven narratives with synchronized speech.
All parameters are passed in the input object of the run request.
| Parameter | Required | Description |
|---|---|---|
| prompt | Yes | Text description (1–5000 chars) |
| aspect_ratio | No | Default 16:9. Options: 16:9, 9:16 |
| resolution | No | Default 720p. Options: 720p, 1080p |
| image_urls | No | Up to 2 frame images (max 10 MB each) for image-to-video |
Veo 3 generates audio natively. Prompts that imply sound — "rain on a tin roof," "crowd cheering in a stadium," "whispered conversation" — produce richer, more immersive output than purely visual descriptions.
Upload a first-frame image and a second image as the end state. Veo 3 interpolates between them, giving you control over both the starting composition and the final pose or framing.
Phrases like "slow tracking shot," "static wide angle," or "handheld close-up" translate directly into camera movement. Vague prompts yield generic motion.
generate_audio toggle — the model does not expose an option to control audio outputNo. Veo 3 Fast always generates audio natively and does not expose a generate_audio toggle to turn it off. (Some other models, such as ByteDance's Seedance, do provide a generate_audio switch.)
No. Veo 3 does not expose a duration parameter. The model determines clip length based on the prompt content. Typical outputs range from a few seconds to around 8 seconds.
The first image sets the opening frame; the second sets the target end state. The model generates video that transitions between them. You can also provide just one image to anchor the starting frame only.