“Dialog” or “Feather's Journey” are short, several -second shots generated entirely by AI, but they already show the scale of the quality stroke.
Further part under video material:
VEO 3 combines diffusion architecture with specialized transformers and – newly new – with an internal audio module. The latter generates sound background, effects and dialogues during the same course, so The image and sound “grow up” together instead of being glued after the fact. Google declares that the model sticks to the prompt much better than previous versions of VEO, and thanks to the Gemini Reasoning layer (Gemini's reasoning) can accept complex commands in natural law.
In the API VEO 3 mode provides Movies in 16: 9 aspect ratio with 24 fps animation, maximum in 720p quality And only up to 8 seconds at one request. The creators emphasize that this is an illustrative version, but this is how the restrictions look today.
Check also: A popular investor shares the vision for the future. We can't keep up with artificial intelligence
Why only eight seconds
The eight -second limit does not kill Internet users' enthusiasm yet. What's more, experienced studies and home -made youtubers are doing a simple, though labor -intensive method of “clips relay”. The script is divided into shots or scenes, describes a coherent set of hallmarks (color palette, camera style, character character), and then generates the segment after the segment. The resulting files are connected in a non -linear editor – exactly as photos from the film set are connected. The camera movement and light can be unified with appropriate filters, and dialogues are improved in post -production. The authors of tutorials point out that a consistent prompt and manual “burn” of frames between segments allow you to maintain the fluidity of the narrative.
Users in 73 countries have already received full access to VEO 3, including in Great Britain, but EU countries – due to the ongoing regulatory findings – were not on this list. Google announces a “quick extension” of availability, but does not give dates.
Therefore, European creators use two gates: “laboratory” accounts (randomly assigned) or VPN masking IPs to American. The latter method allows you to undergo a registration process, but can block payment or reset render loans, because the billing mechanism verifies the country of credit card. In other words: it works, but there are problems and Formally breaks Google AI Pro regulations.
Read also: Is it still worth studying? Artificial intelligence and the future of mental work
VEO 3 opportunities and shadows
For VEO 3 business, it means rapid video production, low -budget ads and content on social media – Without film crews, actors, musical licenses. Already, some agencies sign the clips with the slogan “generated with veo 3” and report several times a decrease in the cost of a minute of finished material.
Personalization is also a chance. The same campaign can go to different groups with a local teacher (and language) and a changed background generated on demand.
Ulianopolis City Hall in Brazil created a complete advertisement from VEO 3, spending only $ 52. for AI loans. The production of a professional quality of a minute advertising film rarely costs less than about 17 thousand. hole. in this country. This amount takes into account the employment of an agency or production company, the whole team (direction, creation, writing, camera, assembly, lighting, sound recording, sound and visual effects), costumes, cast with many actors, copyright, studio rental, stage design and specific elements, such as animals in the stage. Meanwhile For just over $ 50 We managed to generate such material.
The other side of the coin is a field for abuse. The editors of Time alert that VEO can create realistic scenes, e.g. street riots or election counterfeits, and the built -in watermark is easily removed. The fight against disinformation becomes even more difficult, and the governments and social platforms discuss the compulsory determination of synthetic content.
Experts see here a replay from the road that the text generation has passed: Short clips are already at a satisfactory level. The next stage will be Story-Level consistency (consistency at the level of history), i.e. models understanding the entire story – Google tests it in internal projects as part of the Flow service, as well as native to combining scenes into one stream without manual assembly. With the increase in resolution (4K) and the duration of the clip (initially minutes, then quarter of an hour), the need for new copyright and image rights will appear (or at least should).
Don't miss: The EU cannot keep up with the AI revolution. “The EU budget was created in a world that no longer exists”
How to use it today
If you operate in Poland, the easiest way is to sign up for Wait-Lista Gemini Ultra and wait for official unlocking. The alternative is the American VPN plus a payment card issued by a bank from the USA or UK – However, you have to take into account the possible bank of the account. Technically, for the creation of a several-minute film: a script divided into 8-second segments, consistently built prompt, patience for rendering and a classic Davinci Resolve or Adobe Premiere Pro editor. The rest, surprisingly well, will do VEO 3 for you.
There are also alternatives to VEO, such as Runway Gen-3 Alpha, Luma Ai or Pika Labs.
That's enough to Understand why investors and Hollywood look today at the Google model with a mixture of delight and anxiety. If the story teaches that the better the camera, the more movies, then the generative AI camera in the cloud means their unimaginable harvest.
Author: Grzegorz Kubera, Business Insider Polska journalist
