The much anticipated Google I/O 2024 developer’s conference kicked off on Tuesday with the expected slew of announcements covering AI, applications and future tools.
The phrase AI was mentioned over a hundred times on the keynote day alone. But alongside all the Gemini hype, there were some genuinely interesting bits of video and image gen news emerging from Google labs, which bode well for the future.
Note the emphasis on future. Sadly the best of these ‘experiments’ are currently only available on a wait and see basis. Here's how you can access the tools already active and get on the waiting list for those models not yet launched.
Accessing the AI test kitchen
Top of the list is the new VideoFX, the Google text to video alternative to OpenAI’s Sora. The new tool is based on Google DeepMind’s Veo model, and lets the user generate 1080p quality video clips ‘beyond a minute’ in length.
The video clips on the Veo demo page are impressive, and promise some cool upcoming features such as clip extension, video from still image and masked editing.
Unfortunately this new wonder is not available just yet, you’ll need to sign up to the project waitlist to gain access.
The Google text to image model, Imagen 2, launched three months ago, and the tech is already available to use on the ImageFX website.
However adventurous souls who wish to test out the brand new updated Imagen 3 model, will again need to join a...yes trusted tester waitlist.
The current image results are good, but certainly nothing to shout about from the AI pioneer. Here’s hoping the new version offers a significant improvement.
Making music with AI
MusicFX was launched back in December last year, and at the time was a solid attempt at an AI text to music generator.
However five months equals two AI lifetimes, and the tool is now hopelessly outclassed by newcomers like Udio and Suno. And things are probably only going to get worse with the upcoming release of ElevenLabs Music.
Nonetheless Google is fighting valiantly to remain relevant, with an upgrade to MusicFX at this week’s I/O showcase. The new DJ Mode lets you mix various genres alongside your text prompt, using sliders to adjust the intensity of each.
There’s still no vocal action, but the results are good, although again sub-par when compared with the lush complexity of services like Udio. The good news is that both DJ and Creator mode are available now at the AI Test Kitchen location.
A valiant attempt for relevance
Overall it’s been a hard 12 months for Google, which increasingly looks like a flat footed giant caught napping at the AI kitchen table.
While the company is slowly delivering new applications, much of it seems like a desperate attempt to catch up with more nimble and creative rivals.
During one keynote presentation on the power of Gemini 1.5 to deliver multimodal speed to NotebookLM, a tiny sentence at bottom right of the presentation screen declared ‘audio has been pre-generated’. Not a great look for the creators of DeepMind.
