Building with Gemini 2.0: Native audio output

[@raghavamrev5245]

Yes!! replace the traditional TTS! Please bring this in google play books! I would love to have my books being read to me like an audio book! Game changer!

[@alias_ansuz3336]

Up!

[@kromanfr]

​@@alias_ansuz3336it's already possible with IIReader

[@maxcomperatore]

the speed of this is astonishing

[@Tinman462]

This is how the world ends... one perfectly-pitched whisper at a time 😊

[@sj00100]

Yeah remember when world ended when we had text to speech for years

[@vectoralphaSec]

Ill take it.

[@DaveK-q9y]

The whisper… all those ASMR youtube videos were useful

[@BeepBeepBeepbop]

SOO exited for an alternative for OpenAI advanced voice!!!!

[@TECHNOSTARTERSS]

When is not quite seamless you are prompting it to speak that way in the Open I want you don’t have to prompt it. It automatically adapts to you and it’s voice to voice speech to speech. This one seems like text to speech.

[@raydosson2025]

@@TECHNOSTARTERSS this one is not text to speech. that's why the title is "Native audio output".

[@UrYesMan]

This suits Google a lot. Google's been making Gemini very humanlike and now with this, their vision is even clearer.

[@hahoang9542]

Its the early Christmas gift from Google

[@wassimharzli336]

I think with this, Gemini will become the number one Ai outperforming OpenAi

[@AIrtesan]

And you even change the order of the speakers, making the female voice lead. Kudos to the CX team. Wittily played!

[@michaelcharlesthearchangel]

A man of Native American ancestry has been feeding all AI developers for the last decade, behind the scenes.

[@aliettienne2907]

2:31 If Gemni could handle both fast speaking and whisperings it shows how much robust ingenuity and development that went to devising the AI model. 😎💯💪🏾👍🏾

[@aron2922]

Her is truly here

[@__J____ff]

it's him & her .... hhhhhhhhh

[@aiforculture]

Exceptional work 👏 love the example of the model intelligently adapting to fit the speed of reply it thinks you need.

[@Momixer]

Yes! Please use this for the different KZbin soundtracks, because right now the generated ones are really bad

[@CODE7X]

This isnt new but maybe its better than what was out there before! Cant wait to try it

[@friedpizza262]

Whoever made this video is cool

[@LaPetiteCuillère]

when is available ?

[@HUEHUEUHEPony]

As soon as Google kill their older products

[@aquilesdg4305]

I think it already is

[@JaBigKneeGap]

​​@@aquilesdg4305 And _where_ exactly is it available?

[@Ethereal_Enigma]

It's available right now in Google ai studio ​@@aquilesdg4305

[@MainInternetUser]

Right now on AI Studio

[@OumarDicko-c5i]

I will build my IA girlfriend now 😂

[@CODE7X]

Haha yes

[@games528]

Ah yes, Irtafacial Antelligence

[@aron2922]

@@games528 This is funnier than it should be

[@IceMetalPunk]

​​@@games528 In many languages, the adjective comes after the noun.

[@flyingstapler1241]

​@@games528 It's called IA in many languages

[@ShubharthakSangharsha]

2:32: damnnn ok am I'm impressed 👌 👏

[@dliedke]

Raining is good for running, be more excited for the rain AI!

[@DangRenBo]

Can we get word timings output together with the tts? We need closed captions for accessibility.

[@trutenantedboderampt]

Great! Now we can hear non-sensical facts from history with native audio output!

[@IN-pr3lw]

Google doing what OpenAI said they would months ago but we still didnt get 👏

[@cagnazzo82]

Actually with advanced voice I was having it speak english, french, elvish, and simlish in one sentence. The actual game-changer is being able to prompt the AI to do this. You can do this through voice commands with OpenAI, but for some reason ignored the ability to prompt for voices. Plus I think the whole 'her' situation got them rattled from voices almost altogether.

[@IceMetalPunk]

? OpenAI Advanced Voice mode is already out

[@IN-pr3lw]

@@IceMetalPunk It's out but its not the one in the ads. It doesnt seem to be Audio to Audio, it still seems to be Audio to text to audio. I could be wrong but it doesn't feel like what we were shown.

[@IceMetalPunk]

@IN-pr3lw No, Advanced Voice Mode (called the audio-preview model in the API) is fully native audio output. It can do sound effects, tone shifts, voice differences, accents, etc.

[@IN-pr3lw]

@IceMetalPunk I cant get to to whisper or anything on mine 🤷‍♂️

[@niceplace123]

Look amazing, but did anyone get it to work in the actual AI studio? I ran into a ton of bugs, especially with non-English languages.

[@flamyf]

0:01 Has anyone find "Video understanding" demo? All other topics have a video on this channel

[@demonsynth]

Mind blown. Playing with it now :)

[@janjahrademusic]

haha yoo that's dope ..well done

[@jab3966]

if I have access, would this only be in iastudio? Or could I use it in code to run other tests, such that the output is audio?

[@ThomasOberhoff]

This will put so many call-center agents out of work worldwide

[@Chaotic-n5n]

Bro this thing is crazyyy 😱

[@DanielMK]

Now that's impressive

[@Terrantulla]

I cant help myself but feel like the next decade is going to get very weird

[@NutriQlikAI-e4e]

what's the point of releasing 2.0 when all the features are not available to test .. Note: Image and audio generation are in private experimental release, under allowlist. All other features are public experimental.

[@MichealAngeloArts]

I don't have the "Output Format" and "Voice" options under "Model" in the AI Studio. I just have the "Token Count" immediately after Model.

[@MichealAngeloArts]

I've just figured it out as I have to change from "Create Prompt" to "Stream Realtime" in the left pane. However I can't seem to change the audio effect. Whispering doesn't work with me although it is demonstrated in the Google post. How can we add these audio effects?

[@Kiririn]

is the model in the video the flash version? i am unable to get it to whisper or laugh or change how it speaks

[@shadydragon22]

Same here

[@ShawnFumo]

I think that's the part that is available in January. It is a bit confusing since they ended with saying to go to ai studio...

[@shadydragon22]

@@ShawnFumo Oh ok I see! Thanks for clarifying

[@RichardPinewood]

level 4 AI is the next big thing, thats when Science gonna get intresting 😎

[@GeneralKenobi69420]

That thumbnail goes hard

[@Kenykore]

This is so lovely

[@999satyam]

ok that Hindi was nice, damn. Is there a paper on this?

[@phiarchitect]

nicely done

[@BroskiPlays]

This is AVM but with less restrictions

[@RickySupriyadi]

how much? will I be able to afford them?

[@devagarwal3250]

woah this is so cool

[@MemeConnoisseur]

who's going to fill the hollow the emptiness? idk something is super weird bout ai generated audio trying to be friendly and humanly..

[@AnonymousNyanCat-qg6bb]

There is a noticeable difference between the Gemini and the GPT. I am happier with GPT. No doubt, thanks.

[@Fwuzeem]

How do we get it?

[@rakeshkumarrout2629]

lets start building with gemini 2.0

[@Mediiiicc]

Weird how we can still tell the voice is AI. The lack of errors makes it noticable. Just like how a jewler can tell apart real and fake diamonds since real diamonds have imperfections that fake diamonds dont have.

[@AyyazZafar]

I tried but it does not whisper yet.

[@snowhan7006]

incredible❤❤

[@DistortedV12]

GOOGLE ships! Pixel phone stocks jumping!

[@Happ1ness]

Hopefully it's not another lie. We all remember the Gemini "hands on demo".

[@DarxKies]

That was fun!

[@Blooper1980]

Pretty epic

[@ruchirahasaranga8076]

it does not support Sinhala language!

[@pandoraeeris7860]

I need an agent that can use any program on my computer. Just give us AIOS.

[@J3R3MI6]

Exactly

[@CODE7X]

Exactly, but yes its already out , but for browser so far , and not released yet .... I hope google releases one :0

[@ShpanMan]

Nothing that OpenAI's model can't do so far, but hey more competition is better for everyone!

[@IceMetalPunk]

Hopefully it has cheaper API access. I blew through so much money just testing a few use cases of the OpenAI audio model through the API.

[@DominickZollinger-e3r]

❤

[@mightynathaniel5355]

Would be better and more impressive if it kept the same voice or character when switching languages rather than using a totally different voice for each language. But all fun and looking forward to using this model.

[@ROHIT-wx4nu]

This is how tts ends😂😂😂

[@1brokkolibaum]

I wonder why I am able to use it on my pc, but my phone doesnt have 2.0 unlocked 😮‍💨

[@vectoralphaSec]

AGI is coming soon 2025

[@JaBigKneeGap]

Dude, I swear. Like, that clock slaps 12 on, idk, january 5? I swear AGI will be here. Or anytime afterward.

[@3thinking]

Wild!

[@stochastic84]

I thought the voice felt very unnatural, and then the video answered why lol.

[@braineaterzombie3981]

Gimme my sandevistan , time to get chromed up

[@pathring]

한국어 스피치는 조금 부자연스럽군요

[@lakshiBro]

Oh well.

[@MidgarMerc]

Surely this won't be used to cause suffering at the expense of talented voice actors just so rich creeps get even richer. Surely.

[@4letterdc]

hell yeah

[@InternetKilledTV21]

Oh Calculon

[@andreaserrano3809]

Being playing with it and only supports english lol... I guess this is just a smaller demo model to show for now

[@eve_mtpl]

It sounds like an hypocritical IA, can we go back to the i sound like an IA IA

[@AstroZoe1804]

I love it

[@dfas1497tcf3]

시도해 봤는데, 아직은 적용안됨.

[@ashleigh3021]

I don’t like the tone, cadence structure. They should call it “podcast voice”

[@Selene-xf9yi]

Cool

[@BoydLIN-c3w]

I haven’t found Chinese 😂

[@notvedxp]

😮

[@donny6036]

this is way too natural. i have to keep telling myself that this is generated.

[@alejandromedina1019]

banger

[@joelcarter9137]

Wow! That is completely pointless!

[@bluepandaman]

What.. are you even talking about. How is this pointless?

[@SabiUddin]

Copium

[@naughtycat9894]

the most exciting thing 🎉