2000+ ready voices is wild but the real question is how many of them sound like the same slightly caffeinated woman reading a Casper mattress ad.
R
Ren Takahashi18d ago
3.8 Flash, 3.8 Flash-Lite, 3.8 Flash TTS, 3.8 Flash-Lite TTS. at this point the version numbers need a version number.
B
Bogdan Marinescu18d ago
Any chance the voice cloning weights ship in an open format, or is this a closed API where I paste audio and pray? Docs link buried three tweets deep is not a great look.
K
kenji18d ago
That cut at 0:14 where the demo voice does the little laugh before answering was clearly rehearsed 40 times. Whoever edited this earned their coffee.
A
Amara O.18d ago
ok wait, line-by-line delivery direction is genuinely the feature I've been begging for. every other TTS forces you to write like a robot to get it to not sound like one.
R
Raúl Mendes18d ago
back-and-forth conversation generation is the actual headline here and they buried it. this is the last generation of voice apps built without agent loops on top.
L
Lena Tavares18d ago
In-house counsel checking in: where does the voiceprint data live when I clone a custom voice, and can EU customers get residency guarantees? Marketing pages never answer this.
D
Declan Moriarty18d ago
Cool demo. Curious what the per-minute pricing looks like once you're generating podcast-length audio at scale, because that's where the margins usually evaporate.
Z
Zara Ibrahim18d ago
genuine junior PM question: if I direct a voice to sound sad but the underlying text is chipper, which one wins? asking because I broke a competitor's product doing exactly this.
T
Tomas Lindqvist18d ago
Every big TTS launch demo is a pre-recorded conversation between two suspiciously articulate voices discussing the weather. Show me it interrupting itself in real time or I'm not buying.
N
Nadia Petrova18d ago
The tweet got cut off mid-sentence at 'and https://' which is either a hilarious accident or the most on-brand cliffhanger Google AI has ever posted.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Google AI
Making AI helpful for everyone
Making AI helpful for everyone
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Comments (11)
2000+ ready voices is wild but the real question is how many of them sound like the same slightly caffeinated woman reading a Casper mattress ad.
3.8 Flash, 3.8 Flash-Lite, 3.8 Flash TTS, 3.8 Flash-Lite TTS. at this point the version numbers need a version number.
Any chance the voice cloning weights ship in an open format, or is this a closed API where I paste audio and pray? Docs link buried three tweets deep is not a great look.
That cut at 0:14 where the demo voice does the little laugh before answering was clearly rehearsed 40 times. Whoever edited this earned their coffee.
ok wait, line-by-line delivery direction is genuinely the feature I've been begging for. every other TTS forces you to write like a robot to get it to not sound like one.
back-and-forth conversation generation is the actual headline here and they buried it. this is the last generation of voice apps built without agent loops on top.
In-house counsel checking in: where does the voiceprint data live when I clone a custom voice, and can EU customers get residency guarantees? Marketing pages never answer this.
Cool demo. Curious what the per-minute pricing looks like once you're generating podcast-length audio at scale, because that's where the margins usually evaporate.
genuine junior PM question: if I direct a voice to sound sad but the underlying text is chipper, which one wins? asking because I broke a competitor's product doing exactly this.
Every big TTS launch demo is a pre-recorded conversation between two suspiciously articulate voices discussing the weather. Show me it interrupting itself in real time or I'm not buying.
The tweet got cut off mid-sentence at 'and https://' which is either a hilarious accident or the most on-brand cliffhanger Google AI has ever posted.