I think I might have a path to a free software narrator. I think my technique is legal anyway but might need feedback. I've had Claude eyeballing histograms, making changes and then evaluating with Whisper, to generate a replacement for

Say

in my AMOS web player. It's been running all night and it's starting to sound okay. Not great, and not as good as the Amiga's one but it's probably good enough. It's written in typescript but I think it'd be easy enough to convert. github.com/bitplane/narrator-ts

i hve a simple text to speach shell app but havent really tested it much

oh cool. I'll compare and see what can be combined

its meant to use the MOSS-TTS-Nano architecture

oh cool it's a neural model? 🙌 yes totally outclasses my 1985 remake!

it would be good to come up with an API for this so the backend can be changed

Yeah I think a lightweight model for backward compatibility and a next gen one for powerful machines would be ideal

yeah exactly

This is really cool, I'll look into it later in the week. Gonna get this thing pushed to npm and get voice working in my AMOS web player first though

💯

I got the

amos-ts

narrator thing in btw and I'm experimenting with this as a layout. Does it seem reasonable?