[18:48 Thu,22.May 2025 by Thomas Richter] |
Veo 3 is the first video AI to generate native audio – including speech and singing, as well as music and sound effects like noises and animal sounds – all matching the respective video. In addition to generating the dialogue from the text specified via prompt, it is spoken by the characters with matching voices, lip movements, and corresponding facial expressions. ;) Google Veo 3
While previously several AI tools had to be used relatively laboriously for such scenes, this now happens with Veo 3 with just one prompt. Here are some examples:
Veo 3 is currently still a preview, so it is constantly being further developed, but its capabilities and quality are already impressive: It is the first video AI to generate clips in 4K with a length of up to 8 seconds. However, only adult humans are generated – special permission is required for children. Here are some of the impressive examples of Veo 3&s capabilities regarding singing and music: as well as voices:
And sitcoms are no problem for Veo 3 either:
Even human movements like walking or dancing look very good. In dialogues, critical things like direction of gaze, eye contact, and the timely and emotionally appropriate reaction in the dialogue also seem very realistic. Apart from the new integration of sound, the quality of Veo 3 compared to the previous model (which was already better than the competition such as OpenAI&s Sora, HaiLuo or Kling) has also been improved. Veo 3 shows a significantly improved understanding of physics, lighting, shadows and object interactions, which makes the videos look even more realistic. The prompt understanding has also been optimized, so that Veo 3 can adhere even better to the user&s instructions. However, Veo 3 is not perfect either – there are still occasional artifacts or (audio) errors. Veo 3 can be used via Google&s new AI video tool Flow (www.slashcam.de/news/single/Google-Flow--KI-Tool-fuer-Filmemacher---Szenen-erst-19305.html), which gives creatives numerous control options, or via the Google Gemini App. For each prompt, 1-4 different versions are created, from which the user can choose. However, the special features of Flow tailored to filmmakers, such as controlling camera movements, style adaptation using reference images, specifying the first and/or last frames of a scene, and extending the video, currently seem to be available only for Veo 2. ### Veo 3 initially only in the USA – for a price of 250 dollars However, Veo 3 will initially only be accessible as part of Google&s AI Ultra subscription (blog.google/products/google-one/google-ai-ultra/), which costs around 250 dollars per month. The subscription (and thus Veo 3) is currently only accessible within the USA – more countries are to be added soon. Currently, there is a 50% discount on the first 3 months for new subscribers. In addition to the use of Flow and Veo 3, it also includes a number of other functions or high usage limits of Gemini, NotebookLM, the image-to-video AI Wsk (but only with Veo 2), YouTube Premium as well as – certainly interesting for filmmakers – a whopping 30 TB of storage capacity for Google Drive, Photos and Gmail. The Google AI Pro subscription for around 20 dollars (first month free for new users) also includes the use of Flow, but only with the Veo 2 model. According to a user report, Veo 3 should also be available within Adobe&s video AI Firefly, which had already announced that it would support models other than Adobe&s own via API. ### Easier than ever: realistic short films via video AI With the help of Veo 3, you can now easily generate pretty realistic-looking clips of talking people via AI. And from the individual clips of up to 8 seconds, thanks to character and voice consistency and dialogue capability, you can also create entire short films. Here are some example clips generated by users:
The examples nicely demonstrate the revolution that Veo 3 has now triggered: Real AI filmmaking just by prompt, without the knowledge of several tools, is moving into reachable proximity – provided the necessary capital is available.  more infos at bei deepmind.googledeutsche Version dieser Seite: Google Veo 3 vorgestellt - Die erste Video-KI mit voll integriertem Sound |
 |
|
|