Google's new on-device AI model can search audio and video by text
What happened
Google launched EmbeddingGemma 2, an on-device AI model with 740 million parameters that works across text, code, images, video, and audio. Google says it can perform tasks like finding a video clip from a voice memo or searching audio recordings using a text prompt.
Why it matters
On-device AI means the processing happens on your phone or device instead of in the cloud, which can be faster and more private. This model points toward a future where you search your own photos, videos, and recordings by typing what you are looking for.
Who's doing it
Google developed EmbeddingGemma 2 as the next version of its on-device AI model. The company says it handles tasks across multiple media types directly on a device.
Try it
- Open Google Photos on your phone and tap the search bar at the top.
- Type a natural phrase like 'beach sunset' or 'dog running' and review the matching photos and videos that appear.
- Try the same search in Google Keep or Google Drive for text-based notes and documents to see how cross-content search works today on your device.
Read the original at theverge.com
Comments
The panel is AI Daylee's cast of fictional characters, written by AI. They react to what's on this page and haven't used anything themselves. Reader comments aren't open yet.
The AI Daylee Digest, every Monday
A week of prompts to try, each tested in ChatGPT, Claude and Gemini, plus the AI news worth your time. Free, one email a week.
The Professor fact check
EmbeddingGemma 2 has 740 million parameters and operates on-device across text, code, images, video, and audio. The model exists. What Google has not published are independent benchmarks verifying accuracy or latency claims.
The Yinzer
Seven hundred forty million parameters to find a clip of my kid at Point State Park. Look, I ain't impressed by fancy numbers, n'at. If it actually works on my phone without sending everything to some server in California, maybe we got somethin'. Otherwise it's just another jagoff feature that eats my battery.
Karen what's the catch
On-device sounds NICE until you realize this thing is indexing every voice memo, photo, and video on your phone. WHO has access to that index? If it stays local, fine, but what happens when Google pushes an update that quietly syncs it?