At the Google I/O 2025 conference on 18 May, Google made official the release of google veo 3, the third version of its artificial intelligence-based video generator. The announcement comes almost eight months after the launch of Veo 2, and is part of the company's strategy to enhance its AI ecosystem with advanced multimedia capabilities. According to information released by Google, google veo 3 expands on the possibilities offered by previous iterations by combining the generation of moving images with native audio playback, previously absent from consumer models.
The main advantage of google veo 3 is its integrated audio engine. Unlike previous versions, it can generate the entire soundtrack for a clip in a single operation: synchronised dialogue, ambient effects and background music are produced natively, without the need for external editing. This advance is based on the integration of algorithms developed by DeepMind and cross-training with Gemini's language models, enabling greater consistency between visuals and sound. DataCamp.
In practice, google veo 3 is currently available in the United States and reserved for Ultra users of the Gemini application, as well as businesses via the Vertex AI platform. This increased geographical restriction is indicative of a gradual roll-out phase: Google plans to extend access to other regions once it has gathered initial feedback and measured the server load generated by complex requests. Diario ASNumerama.
To make it easier to create films, google veo 3 has been integrated with Flow, the AI filmmaking tool launched simultaneously at Google I/O. Flow combines the Veo, Imagen and Gemini models to offer a visual interface where you can define cast, sets and styles in just a few clicks. Users have access to a complete virtual studio and can generate, adjust or colour-calibrate their sequences directly from a dedicated web dashboard.
The subscription required to use google veo 3 via Gemini Ultra is 250 $ per month (around €272 incl. VAT). This includes a monthly quota of video generation credits and priority access to new features. Businesses wishing to use Vertex AI can subscribe to a professional plan that can be scaled according to the volume of API requests, with pay-per-use billing detailed in the Google Cloud Console. DataCamp.
Google veo 3: technical feedback
Initial technical feedback highlights an unprecedented level of realism. Tests carried out by the specialist press show that google veo 3 produces fluid transitions and remarkably precise lip-synchronisation, even on complex sequences. Some of the examples unveiled during the public demonstration were so convincing that it was difficult to distinguish the images generated from traditional footage. The Verge.
However, this power raises major ethical issues. The ability of Google Veo 3 to empower the creation of convincing scenarios raises fears of a proliferation of deepfake content or misleading clips likely to fuel disinformation. Google has strengthened its safeguards by banning certain prompts of a political or violent nature, but the boundary remains fragile when it comes to malicious use. The Verge.
Google veo 3
At a time when OpenAI, Runway and other players are announcing a growing number of video models, google veo 3 is redefining the benchmark for integrated audio-visual rendering. Its gradual deployment in new markets and its possible democratisation via consumer interfaces could disrupt traditional creative channels while opening up new avenues for automated audiovisual production.
