Model positioning
Veo 3 is a video generation model released by Google that can generate videos with synchronized audio from prompts, intended for video creation and visual storytelling. (blog.google)
Capabilities and uses
- Supports text and image prompts and can turn short stories in prompts into video clips.
- Can generate ambient sound, sound effects, and character dialogue; Google also emphasizes its physical realism and lip-sync capabilities.
- Can be used to conceptualize video content, rapidly iterate on creative ideas, and create cinematic narratives. (blog.google)
Limitations and availability
- Generated results are not always accurate: Google notes that natural, coherent speech and audio-visual synchronization in the Veo family still have room for improvement.
- Veo 3 outputs contain SynthID watermarks. Specific features, usage limits, and available regions depend on the access platform and version.
- At launch, Google AI Ultra subscribers in the United States could use it in the Gemini app and Flow, and enterprise users could use it through Vertex AI. The Vertex AI public preview opened to all Google Cloud customers and partners on June 26, 2025. (deepmind.google)
Release date
On May 20, 2025, Google released Veo 3 in its official I/O 2025 announcement. The release identifier in the input is veo-3-preview; this identifier should not be treated directly as a verified API model ID. (blog.google)
Pricing
- Historical preview price, not current pricing: The Gemini API paid preview price announced by Google on July 17, 2025 was 0.75 US dollars (USD) per second, charged for generated video and audio output; the announcement did not separately list a text input price. This price should not be considered valid on the May 2025 launch date or currently. Query date: November 18, 2026. (developers.googleblog.com)
- Current separate input and output prices for the
veo-3-previewrelease identifier in the input: unverified.