Veo
In plain wordsGoogle's video-generating AI, which since 2025 creates short clips complete with matching sound and dialogue from a written prompt.
A series of text-to-videotext-to-video model, n.: An AI that creates a short video clip — increasingly with matching sound — from a written description, such as 'a paper boat sailing down a rainy street'. and image-to-video models from Google DeepMind. Veo was announced at Google I/O in May 2024, with Google claiming 1080p clips longer than a minute; Veo 2 (December 2024) added up to 4K resolution and more realistic physics. Veo 3, released at Google I/O on 20 May 2025, generated synchronized audio — dialogue, sound effects and ambient sound — together with the video, helping to make AI clips with talking characters a viral genre. Veo 3.1 followed in October 2025. Veo powers Google's Flow filmmaking tool, which assembles clips of about eight seconds each into longer scenes with consistent characters, and it is offered through the Gemini app, YouTube Shorts and Google Cloud. Google describes Veo as a latent diffusion modeldiffusion model, n.: An AI that learns to create images, sound or video by practicing noise removal: shown pictures buried under static, it learns to clean them up, then builds new ones starting from pure static. and marks its output with SynthIDSynthID, prop. n.: Google's system for hiding an invisible watermark in pictures, video, audio and text made by its AI tools, so they can later be recognized as AI-generated. watermarks. In May 2026 Google added Gemini Omni, a conversational video generation and editing model; Veo continued alongside it, though as of October 2026 the Gemini API was retiring its Veo models in favor of Gemini Omni Flash.
Veo 3 gave the street interview real crowd noise, and the lip-sync held up.
Origin
Developed by Google DeepMind and announced at Google I/O on 14 May 2024, succeeding Google Research video models such as Imagen Video and Phenaki (2022) and Lumiere (2024).
Related
- text-to-video modeltext-to-video model, n.: An AI that creates a short video clip — increasingly with matching sound — from a written description, such as 'a paper boat sailing down a rainy street'.
- SoraSora, prop. n.: OpenAI's video-generating AI, unveiled in 2024 with startlingly realistic clips; a 2025 version added sound and a social app, but OpenAI shut the service down in 2026.
- ImagenImagen, prop. n.: Google's family of AI image generators, first shown in 2022, whose research showed that better understanding of language yields pictures that match descriptions more faithfully.
- SynthIDSynthID, prop. n.: Google's system for hiding an invisible watermark in pictures, video, audio and text made by its AI tools, so they can later be recognized as AI-generated.
- diffusion modeldiffusion model, n.: An AI that learns to create images, sound or video by practicing noise removal: shown pictures buried under static, it learns to clean them up, then builds new ones starting from pure static.
- Nano BananaNano Banana, prop. n.: The nickname, later Google's official branding, for its Gemini image models, a viral hit in 2025 for editing photos from simple instructions while keeping people looking like themselves.
Sources
- Google DeepMind (2025). Veo 3 Model Card. storage.googleapis.com/deepmind-media/Model-Cards/Veo-3-Model-Card.pdf
- Google (2026). Gemini deprecations. Gemini API documentation. ai.google.dev/gemini-api/docs/deprecations
Mentioned in
text-to-video modeltext-to-video model, n.: An AI that creates a short video clip — increasingly with matching sound — from a written description, such as 'a paper boat sailing down a rainy street'., ImagenImagen, prop. n.: Google's family of AI image generators, first shown in 2022, whose research showed that better understanding of language yields pictures that match descriptions more faithfully., SoraSora, prop. n.: OpenAI's video-generating AI, unveiled in 2024 with startlingly realistic clips; a 2025 version added sound and a social app, but OpenAI shut the service down in 2026.