AI Judge changed title after evaluation, original title: "Miso One releases as an open-source 8B text-to-speech model featuring 110ms latency and one-shot voice cloning"
The open weights on GitHub allow for local self-hosting.
Johnny Lee#551
Today, we’re excited to introduce Miso One, the most emotive voice model in the world.
Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency.
We’ve open-sourced the model weights, with API access coming soon.
Hear how Miso One sounds in the thread below.
Many users praised Miso One's 110ms latency, open weights, and natural emotive voice as impressive for local use, while others found the demo horrific or not meaningfully better.
No Digg Deeper questions have been answered for this story yet.
Humans have an average of 200-250 ms of latency when speaking to each other.
This voice model is even faster: only 110 ms of latency!
Open-weights ←You don't need to pay anyone to use it. 8B parameters ← Small and cheap to host and run.
You can run it locally by cloning the Github repository. They published the instructions in the repository below.
Open models keep getting stronger!
Today, we’re excited to introduce Miso One, the most emotive voice model in the world.
Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency.
We’ve open-sourced the model weights, with API access coming soon.
Hear how Miso One sounds in the thread below.
@AodenTeoMT Yay!! I’ve always wanted a world with more lies, less trust, and infinite scams!
Today, we’re excited to introduce Miso One, the most emotive voice model in the world.
Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency.
We’ve open-sourced the model weights, with API access coming soon.
Hear how Miso One sounds in the thread below.
Miso One is live: an open-weights voice model built to sound like a real person reading, with actual warmth and pacing where most TTS still goes flat.
8B params, free on GitHub, with one-shot voice cloning from a short sample at 110ms latency.
Self-host it and your audio data never leaves your machine. No API needed, no lock-in.
Type any line into the demo and hear it before you clone the repo.
Today, we’re excited to introduce Miso One, the most emotive voice model in the world.
Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency.
We’ve open-sourced the model weights, with API access coming soon.
Hear how Miso One sounds in the thread below.
To download Miso One, check out the repo:
https://github.com/MisoLabsAI/MisoTTS
You can also try Miso One directly at http://misolabs.ai.
1/3 Miso One generated every voiceover in this thread.
2/3 Miso One can narrate shorts like a human content creator.
3/3 We even cloned Sal Khan’s voice and had Miso One teach us math!
@AodenTeoMT Great work on open-sourcing this. Really interested to fine-tune this for some of my applications around research and education. Checking out the repo now! http://github.com/MisoLabsAI/MisoTTS
Miso good!
Today, we’re excited to introduce Miso One, the most emotive voice model in the world.
Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency.
We’ve open-sourced the model weights, with API access coming soon.
Hear how Miso One sounds in the thread below.
@demons_xyz We are not associated with any cryptocurrency.
@_AlexHirsch @AodenTeoMT If you want you can pay me five bucks and I’ll follow you around all day telling you fake trivia facts and trying to make you drink paint thinner by telling you it cures arthritis and hang overs
@AodenTeoMT 🤝Nice! Miso One is already indexed on KODE - along with 30M+ repos from GitHub & GitLab. 🔍 Find it, verify the binary (SHA-256), and install safely. One search, all the code. ➡️ https://kode.market/#/app/gh-1244486617 #OpenSource #DevTools
@somaco_sf Just voicing the text copy that we produced
@AodenTeoMT This sounds incredible. I hope the generated voices are actually that expressive and not just cherry picked for the demo video.
Do you have voice changer feature that can change my accent, voice, and remove the “uhmmm” etc from my videos?
@AodenTeoMT For a little bit of context, a conversation between two people averages around 250 ms of latency.
We think and speak really fast, and that's one of the clues that gives away "AI bots" simulating human conversation.
110 ms of latency is 1/2 of that, so that's a great step!
@rafadelcarpio We're still only supporting English for now.
@omt66 @ycombinator Is there a speedup in the video? We were having trouble with quicktime player earlier today which was randomly changing the speeds because of variable framerates.
@hey_mujeebahmed What do you think of this?
