Yesterday he had Google as protagonist, and between all the news that presented was I see 3. His artificial intelligence generative text to video advanced on the right to all its competitors, being the first capable of generating audio next to the video itself.
With the arrival of I see 3, I see 2 becomes an accessible version. So much, that you can try it in Gemini’s own app in both iOS and Android. In Xataka we wanted to test it, and the results are … good, very good.
How to access See 2. I see 2 now part of the Google AI Pro plan, for 21.99 euros per month. This includes access to Gemini’s model 2.5, I see 2, flow, notebooklm, gemini in apps such as Gmail or Google Docs, and 2 TB of cloud storage.
Access to see 3 will require the 250 dollar planfrom what I see 2 automatically becomes the most mainstream option and accessible to users that do not Look for professional uses with this AI.
In iOS and Android. If you have the Gemini Pro Plan, as soon as you open the app on Android you will now see a video option, next to those of Research and Canvas. Just press this option to access the mode of video creation, with a generation of up to eight seconds.
In the case of iOS, we have to go to the upper section of the app, and change the model that we have selected (2.5 flash, 2.5 pro or deep researc) to I see 2. Once you are configured thus, we can introduce the prompt.
The instructions. Using I see 2 in Gemini is quite easy: we just have to be as descriptive as possible. The more detailed the description, the better the result will be.
For its correct operation it is not necessary to write the Prompt in English, it works perfectly if we give the instruction in Spanish.
Testing I see 2. One of yesterday’s great memes had to do with a challenge that is still choking many IAS: that of a person eating Spaguetti. This benchmark of the absurd implies complex understanding of textures, movement, deformation … and yes, I see 2 solves it. Not perfectly, but solves it.
It is not a perfect. In a petition as simple as the “drone flight about Paradisiaca Beach. There are people surfing, palm trees, floring seagulls” has generated a fairly poor result. An idea to improve it? Having asked him not to be people, since generating so many elements has distorted the result.
In this last example, two people dancing flamenco with a black background, surprise hair quality, how it has taken care of makeup, gestures … having made such a close plane the dance is not well appreciated, but it is another proof of how this AI works.
There is no open bar. There are no more examples not because I don’t want to, but because Google has not left me. After generating a few videos (3,4) he has told me that he could not generate more until after half an hour. After half an hour … he hasn’t left me either.
It is clear that paying 20 euros is not enough to create infinite videos, so we will have to use this tool with a certain moderation.
Image | Xataka
In Xataka | Google first taught its new glasses with a real demo: an unforeseen live and a bet that points away
GIPHY App Key not set. Please check settings