Meta has released, SeamlessM4T, a new AI model for transcribing, translating, and speaking languages. It’s really impressive.
We’re finally getting close to a babble fish!
What’s really neat is that it can handle mixing languages. I was able to record a sentence where I switched back and forth between English, Spanish, and Portuguese and it figured it all out.
Can it run in-device?
Released meaning open source?
It is open source!
https://github.com/facebookresearch/seamless_communication
And YES, you can run it on device.
https://github.com/facebookresearch/seamless_communication/blob/main/docs/m4t/on_device_README.md
And you can play with it on Huggingface!
Wow, amazing how well this can transcribe audio, even when I'm talking really fast and mumbling.. would be hella cool to have real time verbal conversations with my IDE
I think the quality of the translation will be poor. I don't speak English well and rely on a translator for my writing. The accuracy of AI translation has a long way to go.
