Show HN: MARS5, open-source, insanely prosodic TTS model (github.com)

Hey guys,

This is Akshat from CAMB.AI. Today's we're super pumped to introduce MARS5, a fully open-source (commercially usable) TTS with break-through prosody and realism available on our Github: https://www.github.com/camb-ai/mars5-tts

Watch our release demo here: https://www.youtube.com/watch?v=bmJSLPYrKtE

Why is it different? MARS5 is able to replicate performances (from 2-3s of audio reference) in 140+ languages, even for extremely tough prosodic scenarios like sports commentary, movies, anime and more; hard prosody that most closed-source and open-source TTS models struggle with today.

We're excited for you to try, build on and use MARS5 for research and creative applications. Let us know any feedback on our Discord: https://discord.gg/ZzsKTAKM

7 comments

[ 4.5 ms ] story [ 25.7 ms ] thread

[dead] NourMerey 2y ago ↗

[flagged]

bosky101 2y ago ↗

Wow! Well done

satvikpendem 2y ago ↗

Can I use this to convert ebooks to audiobooks? What's the amount of RAM and VRAM I'd need?

arnavmehta7 2y ago ↗

Works like a charm for me on a T4, faster on L4. Basic free-tier GPU enabled colab notebook is enough to get started.

fivestones 2y ago ↗

With a free tier colab notebook (T4), it takes about 1-2 minutes per sentence for me. Am I doing something wrong? At this rate converting a whole ebook to an audiobook seems like a no-go at least for the free tier. I’m using the deep clone. Any tips for making it faster?

wanderingmind 2y ago ↗

Great work. Would it be possible to use the model with other tools like piper that are more lightweight?

fivestones 2y ago ↗

This seems pretty unlikely but I’d love to be able to do the same if someone knows a way to