Gemini 3.8 Flash TTS with Voice Cloning

Author: Sam Witteveen · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Gemini 3.8 Flash TTS with Voice Cloning · Channel: Sam Witteveen · 17:02 · Watch on YouTube

Topics: AI evaluation and benchmarks, Multimodal models

Summary

Sam Witteveen reviews Google’s Gemini 3.8 Flash TTS and Flash Lite TTS, comparing their intended uses, features, benchmark results, pricing, and generated audio. Flash TTS is positioned for expressive, creative work, while Flash Lite targets higher-volume generation where cost matters. The video demonstrates voice design, speech-style directions, inline vocal tags, multi-speaker dialogue, and voice cloning with a consent check. The reviewer finds the voice-design controls useful, but says independent benchmark results are more mixed than Google’s claims and that the cloned voice sounds noticeably different from the original.

Key points

Flash TTS is aimed at expressive applications such as character voices, audiobooks, podcasts, and games; Flash Lite is intended for bulk generation, dubbing, and voice agents.
Users can describe a voice in plain language, including its role, accent, and characteristics, and can choose from a library of more than 2,000 voices.
Voice cloning uses a 30-second reference sample and requires a spoken consent clip that must match the reference speaker.
Generated clips include SynthID watermarks and C2PA credentials; voice cloning is unavailable in some countries and U.S. states.
Speech can be directed line by line, with styles such as whispered, rushed, or sarcastic, and inline tags for sounds such as sighs, laughs, and gasps.
Google says multi-speaker scenes can maintain consistent voices over long-form audio, addressing voice drift that the reviewer has encountered in some open systems.
On Artificial Analysis preference ELO, Flash TTS ranks second and Flash Lite ranks sixth; the reviewer notes that some of Google’s headline benchmark claims rely on Hume AI benchmarks.
The reviewer says Gemini performs particularly well on pronunciation and multilingual speech, while Google’s pricing sits around the middle of the compared offerings.
In the demos, the reviewer finds Flash Lite somewhat lazier than Flash TTS and considers its price difference relatively small.
The reviewer says voice design is a strength of Gemini, while the demonstrated voice clone sounds quite different from the original recording.

Chapters

Start	Chapter	Summary
00:00	Introducing the Gemini TTS models	The reviewer introduces Flash TTS and Flash Lite TTS, their intended uses, and Google’s claims about expressiveness and benchmark performance.
02:08	Voice design, cloning, and safeguards	The video covers customizable voices, the voice library, consent-based cloning, safety measures, and line-by-line performance directions.
04:45	Benchmarks, languages, and pricing	The reviewer examines benchmark sources and independent rankings, then discusses pronunciation, multilingual performance, and relative cost.
07:22	Testing the models in AI Studio and Colab	The reviewer demonstrates voice prompts and compares Flash with Flash Lite, including style directions and inline vocal tags.
11:37	Voice design and multi-speaker dialogue	The demonstrations show how a described character voice carries across speech and how two speakers can be used in one dialogue.
13:21	Voice cloning demo and final assessment	The reviewer explains the reference and consent recordings, compares a clone with the original voice, and gives a final assessment of the models.

Quotes

Google’s pitch is that these are the most expressive voice models yet — 00:30

the person whose voice it’s gonna actually create has to go through a spoken consent clip — 03:11

I’m not sure if quoting benchmark numbers from a benchmark that you designed yourself is the best way to judge this. — 05:19

Okay, so that one certainly is flat and bored. — 10:02

For me, it sounds quite different. — 15:26

Concepts and entities

Gemini 3.8 Flash TTS, Gemini 3.8 Flash-Lite TTS, Google DeepMind, AI Studio, Hume AI, Artificial Analysis, Qwen TTS, SynthID, C2PA, SoundStorm

Full transcript (original language)

00:00 Okay, so I think it’s fair to say that Google has kind of dropped the ball on LLMs lately. The proposed Gemini 3.5 Pro never came, and hopefully all of that’s gonna change very soon with the Gemini 4 release. But on the multimodal side, it’s a very different story. The DeepMind multimodal team have been doing really well with the multimodal releases that have been based on the Gemini Flash models. So we’ve had Omni 1.1 Flash update. On the voice side, we’ve seen the live models with things like

00:30 live translate, and subscribe. And now there are two new text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash Lite TTS. Now, Google’s pitch is that these are the most expressive voice models yet, And they’re number one on a bunch of the benchmarks. and certainly some of that holds up, but when you look at some of the independent numbers, It’s perhaps a bit more mixed than the blog post makes it sound. Okay, so first off, there are two models. They’re aimed at different jobs. Flash TTS is the creative one.

01:00 This is for things like character voices, audiobooks, podcasts, games, all those kinds of things. flash lite TTS on the other hand, is the volume one. It’s built for dubbing, bulk audio generation, voice agents, and anything where you’re generating a lot of speech and cost matters more than getting every single nuance perfect in there. so the headline feature here is all about voice design. instead of just picking from a list of preset voices, you describe the voice that you want in plain language, you give it a role, an accent, the

01:34 characteristics of the voice, and it builds that voice for you across over 100 different languages and this is something that we’ve seen both from other TTS providers, but also from open models. But certainly does seem that Google’s improved the whole concept of this here this is one of the things that I found often to be the most frustrating with TTS models is it while things have been getting fantastically better and we’re starting to see really good prosody in the models, we’re starting to see really good human likeness, et cetera. Often the challenge becomes is that when you go through the test voices

02:08 that they’ve got, none of them really fit what you’re trying to do And so you end up having to try and do some kind of voice cloning. and obviously the challenges there can be a lot of legal issues. So we have seen some really nice attempts at this from the Qwen TTS people. but some of the more recent ones that I’ve looked at No matter how hard I try to describe the voice, it ends up going in a slightly different direction In fact, what I’ve often ended up doing is basically generating a lot of different voices with slightly different descriptions,

02:38 going through and curating the best one, and then actually cloning that artificially made voice now alongside voice designing here, there’s a library of over 2,000 different voices in here. They’ve got a whole bunch of different regional accents and you can actually sort of take a voice manipulate it the way you want, and then actually save that voice so that you can use it going forward. And I guess this is really no surprise. Google’s had this feature in testing for quite a while, but they just haven’t rolled it out, And that’s the whole feature of voice cloning. So you give it a 30-second sample and it creates that voice.

03:11 Now obviously compared to other startups, when Google does something like this, they get in way more trouble if it goes wrong. so they clearly have taken their time with doing this, and they’ve really focused on putting some guardrails on this So the person whose voice it’s gonna actually create has to go through a spoken consent clip, and that clip has to match the reference speaker before the clone ever gets created. on top of this, of course, they’ve also got the whole sort of safety features in there that every clip these models generate is watermarked with SynthID, and

03:41 there are C2PA credentials on top of that. the other challenge that a lot of people are gonna face is that because of legal requirements, that voice cloning in AI Studio isn’t available in a lot of different countries, and even in some states in the United States So it does show just how touchy of a subject that can be, especially for a big company like Google. Now once you’ve actually got your voice, the other big thing you can do here is you can give stage directions line by line. So you can say that this one should be whispered, this one should be rushed, this one should be sarcastic, et cetera.

04:13 And you can drop in sort of non-verbal tags like I’ve shown you in some of the other TTS systems for things like sighs, laughs, gasps, all those kinds of things another nice feature that you’ve got in here is you can actually do two speaker scenes from a single script. And Google claims with that, that it can actually hold the voices for hours of long form audio without the voice sort of actually drifting. That’s one of the issues that I’ve run into with some of the open TTS systems recently, is that once you get that voice right and you start

04:45 generating anything of any length, it starts to end up drifting off. Alright. So let’s talk a little bit about the benchmarks because this is where it gets interesting, but also a bit questionable in some of the things that they’ve done. So Google’s numbers mostly come from Hume AI. They talk about that the Flash TTS is number one on Hume’s voice design benchmark, and it leads on accents and you can see people like Logan actually tweeting this out But here’s the thing. One of the two people that wrote the blog post is Alan Cowen. And Alan actually was the CEO and chief scientist at Hume until January this

05:19 year When Google DeepMind licensed their tech and hired a bunch of the people from that startup over. So I’m certainly not saying that their numbers are wrong But I’m not sure if quoting benchmark numbers from a benchmark that you designed yourself is the best way to judge this. Now, the good thing here is that artificial analysis have also done their own tests on that And I think those are reasonably independent if we look at the sort of preference ELO, which is the common way that people actually test these kind of models. You can see that the Gemini 3.8 Flash TTS is coming second in here

05:50 And the flash lite TTS is coming in sixth in here And we can see that the flash model is just ahead of the new Qwen Audio 3, TTS Plus Now, at the time of me recording this, that one is still not an open model. But if Qwen goes on to release the weights for that as an open model like they did with their Qwen TTS before, that’s gonna be really interesting to see as well. Where the Gemini models really do seem to take the lead are some of the benchmarks around pronunciation et cetera. that’s especially useful if you’re using the flash lite model to bulk generate

06:21 audio of where you’re saying things like numbers, IDs, things that people really want to be able to hear correct the first time That’s something that really does matter if you’re building a voice agent with this kind of model. if we look at how the models actually do on other languages, we can see this is, and I wouldn’t argue that this is Google’s strength, right? Google has a whole bunch of different multilingual data. and you can see this clearly in here of looking at how they compare to other providers. that said Cartesia and ElevenLabs here are really making a strong move at the

06:51 multilingual TTS systems finally the other big demo here is actually price. So Google’s kind of sitting in the middle of the pack here, they’re obviously a lot better than some of the offerings from Cartesia and ElevenLabs, and certainly from Minimax, et cetera but they’re by far from being the cheapest in here. And I think in some ways more importantly, the flash lite model isn’t hugely cheaper than the flash model itself Which is definitely a difference than when we compare the text versions of these models. All right, so Let’s go over to AI Studio, have a play with this, and then we’ll

07:22 try out the two models, and then we can try all this out in Colab as well all right, so if you want to check out the voices, you can actually come into AI Studio, and just pick one of the voice models. You can see here I’ve picked the Gemini 3.8 Flash TTS. and then you’ll be asked to basically hook up, an API key to it. but from there you can basically go through and you can explore a whole bunch of different models in here So if I just come in here, I can basically take the model, and then I can make a description in here. And the cool thing is if you want to, they’ve actually

07:53 got voice rewriting in here. So you can see that if I click on that, that will basically just use Gemini to flesh out, the actual, voice, prompt in here of what I want the voice to be like. And then you’ll see from there it will give you three voice options that you can use. Everything from describing the person themselves through to, the position that they are on a microphone and that kind of thing. So that’s really kinda cool Unfortunately, as I’m recording it now, it seems like the voices are failing in here.

08:24 But normally what you would do is you would get basically three voices, and those voices then you can customize it, you can then refine it, et cetera right after multiple attempts to try and generate this, it doesn’t seem like this part is working now. Let’s just jump into the code and have a play with the actual Colab for this All right, so let’s take a listen to it Hi, I’m Sam, and today we’re looking at the new Gemini 3.8 text-to-speech models So as you can see, the, the basic voice just works pretty well. and the cool thing is you’ve got lots of choices if you wanna do just, sort

08:56 of standard voices and stuff in here. let’s come down and listen to the comparison between the flash model and the flash lite model The quarterly numbers are in. Revenue is up 11%, and churn is the lowest it has been in two years Okay, so that was the normal flash TTS. Now let’s listen to the flash lite The quarterly numbers are in. Revenue is up 11%, and churn is the lowest it has been in two years. Okay, so the common thing I notice with the flash lite TTS is

09:28 it tends to be a bit more lazy. it might go on for longer, et cetera. And honestly, I feel that the price difference is not huge. So you’ve got the Flash TTS being 50% more. that if I was doing something that wasn’t too long, I’m probably just gonna go for the good voices Okay, so you can see now we’ve got the style control. This is basically where we can give it direction on how we want, what we’re asking it to speak, how we want it pronounced, et cetera We just got the results back, and the prototype actually works

10:02 Okay, so that one certainly is flat and bored. Let’s try thrilled and breathless We just got the results back, and the prototype actually works Okay, and then now nervous, almost whispering. We just got the results back, and the prototype actually works. Okay, so you can hear that all three of them definitely sound very different, even though they’re the exact same words. So one of the things that I do like about this, is that this affects the length, of the audio. So with some of the open models, one of the challenges that you have is

10:34 it will try and say something in the style that you’re asking, but it’s also trying to hit a certain time length, in the way that it returns back. Here, we’re basically letting the time length be dictated by the style that it’s saying. All right, The other thing that you can do in here is that In the actual script that you want them to say, you can actually put in, these inline tags. And so these are things like whisper, laugh, sigh, that kind of thing. Wait. Shh. Did you hear that in the hallway?

11:04 Never mind. It was just the cat. You should’ve seen your face Okay. So you could hear there that the inline tags, it certainly, does what they’re basically there for. In the case of whisper, sigh, laugh. The laughs will change also depending on the style directions you’ve given it as well All right, so next up is the whole idea of voice design. So now what we’re doing is we’re not just passing in, a script and sort of a guidance of how to speak it. We’re actually giving a description of the voice itself.

11:37 So the idea here is that this voice should then carry over to everything that it’s gonna speak, even if we give it other sort of descriptions of how to speak it going through. So this is where you can put in things like, where the person comes from, the accent that you want, if you want to define something about age, that kind of thing in here. so you can see in here I’ve got a character called the astronomer, that’s in his late 60s with a gentle British accent

12:14 Okay, so you can see there it’s got some vibes of, sort of a David Attenborough voice there, but without actually mentioning him personally, I feel the voice sounds a lot younger than late 60s, but let me know in the comments what you think if you try that out All right. Next up is the fun one where you can basically have multi-speaker dialogues going on in here. So here we’re doing one API call with, two voices. and so we’ve got this idea of, per turn style, Meaning that we can basically

12:45 set what each character says in here, and you can see we can define the actual speakers, who they are, but we can also define the style for each line of dialogue that they’ve actually got. So I’m not gonna play the whole thing here. I put the, notebook up. You can have a play with this yourself and try it out, and just see how it goes. But let’s take a little listen to this

13:21 Okay, you can hear there this has definitely got the vibes of sort of a NotebookLM, podcast kind of thing that gets generated. Now for me, I don’t know if it’s just that I’m so used to hearing these voices now that they kind of sound fake to me. But the cool thing is you can swap out other voices. You could make your own, et cetera, as you go through this all right, lastly, we’ve got the voice cloning in here, and the voice cloning can be done, just by code, right? So you can see the code here is basically just making an HTML interface where

13:53 you can record, the reference audio, so which is gonna be what you want the audio to sound like, and then second, where you can record, the consent audio. I do think that Google’s in a very hard spot here. We know that this is something that their TTS systems have been able to do for quite a number of years now. and this goes back to a paper called SoundStorm, which, at the time was, absolutely amazing. This is more than three years ago now And even then they were able to basically clone voices so well.

14:23 and at the time I was actually given a demo of this, And it was amazing just how well it could clone someone’s voice back then jump forward to today, many of the open models now have got this worked out. And yet this is the first time that we’ve seen Google actually make this available to people. And I think it’s been very clear that Google’s very conscious of the ramifications of this, and that people can use this kind of tech for doing a lot of nefarious kind of things But I got to say, in some ways, I wonder whether Google’s version here

14:53 has actually been, dumbed down a little bit to be not quite as good as some of the open versions that are out there. so basically what you do here with this HTML interface is that you can, record in your reference audio, so this is the audio to replicate, and then you can record in the consent audio. and this is the really important bit where you’re basically giving Google, consent to make a synthetic voice based on your voice And if we come down and actually listen to these, first I’ll play you the original, bit that I recorded in here

15:26 The cosmos is within us. We are made of star stuff. We are a way for the cosmos to know itself. Okay, so that was the sort of original recording I did, and now let’s listen to the clone of my voice done with this This was generated from a 30-second sample of my voice Now I’ll leave it to you guys to decide. For me, it sounds quite different. I think possibly had I uploaded a longer reference audio, it might be able to get a better version of my voice. Probably the best model that I’ve seen doing this, open wise was the Qwen TTS.

15:59 I made a whole video about that when it came out, and I still think that one is really amazing at how well It can clone voices. And in some way I would say that was their strength, at least back then. and yet They were probably weaker with the descriptions for creating a voice. and that’s the Gemini’s strength in here now. So just to finish up, I would say this is definitely worth checking out. It is really cool, how easy it is nowadays to basically do voice things. Things that just used to be so difficult, in the past are now a very simple API call.

16:31 And I think for people who perhaps don’t have hardware to run their own TTS system, to run something locally, This is gonna be quite nice for just giving you an API that you can ping and get high quality voices out with really nice prosody and with a custom style that you can define all by yourself So anyway, let me know in the comments. I know there are TTS models coming out all the time. I try to keep up with them. I’ve missed some. if you’ve got any recommendations that you would like me to cover, please put them in the comments below.

17:02 I’d love to hear that from everyone Anyway, as always, if you found the video useful, please click like and subscribe, and I will talk to you in the next video. Bye for now