Google Rolls Out Gemini 2.0: What's New?
Hot off the press, Google just introduced Gemini 2.0, its freshest brainchild in the AI lineup.
Gemini 2.0 takes a leap forward packing the power to create text, visuals, and sound - a big step up from the older models.
 |
| Gemini 2.0 and the Rise of Google's Live API |
Unpacking the Fresh Additions to Gemini 2.0
Zooming in on the upgrades, Gemini 2.0 flexes its muscles with:
- A new trick to whip up pictures and tunes. Now, Gemini 2.0 can whip up mixed media like images and sounds and tweak them as you wish. Got a photo or video you're curious about? Shoot, and Gemini 2.0 will hit you back with answers.
- It creates vocal sounds in eight unique styles fitting various languages and ways of speaking. You can even tweak how quick or slow it talks. For kicks, you might get it to sound like a pirate.
- It's better at connecting stuff 'cause it can handle pics, vids, and tunes now.
- It's pretty smart with Google stuff and can run code with cool tools like Gemini Code Assist.
- It's a whiz at digging into info longer stuff, and making sure whatever it comes up with is spot-on and super relevant.
- Google AI Studio and Vertex AI now host the Gemini 2.0 Flash beta giving devs a chance to try out its cool new bits and pieces. But gotta say the folks in the early access club get to play with generating images and sounds. Looks like everyone else gotta wait until January 2025 to get their hands on that.
Google didn't stop there - they went ahead and dropped the Multimedia Live API so now coders can whip up apps that use audio and video on the spot.
Gemini's Multimodal Live API is Here
Alright, Gemini's Multimodal Live API is all about letting builders make stuff where words, tunes, and clips all come together. It reacts super fast and smooth. The tech behind it? WebSockets - that's what’s making the magic happen for a quick back and forth with any wait time.
Live API Features
- Back and Forth Chat: People can chat using text, blab into a mic, or hit up a video call all at the same time.
- Speedy Replies: The system starts to spit out answers super quick, like in less than a second (600 ms to be exact), so talking to it feels natural.
- Talk Like a Person: The setup has got a bunch of peppy human-like voices. You can even put them on hold and then get back to the convo, which is pretty cool.
- Getting Video: It's smart enough to get what's going on in videos, which means it can chat about them more.
- Mix-and-Match Tools: You can throw a bunch of different functions together when you ask for something, so you can get complicated stuff done without a sweat.
- Voices You Can Change: You get to pick from five unique voices (they're named Aoede, Charon, Fenrir, Kore, and Puck) depending on what you need.
How to Use
- Getting Going: WebSockets help set up a link where we lay out the model, configure the settings, and make clear the system orders.
- Shipping Data: People can shoot over their chatter talking, and film clips to the Server.
- Hearing Back: The server shoots back answers on the fly, and they might be made up of written words, sounds, or commands to get certain jobs done.
Additional Resources
- Real Practice: Have a go with the tool on Google AI Studio.
- The Lowdown: Wanna know more? Just hit up the Gemini API docs.