messing around with AI anime dubbing. the automatic process is inconsistent with voices and has some weird accidentals like inserting the word "pause" several times at one point, which could all be rectified in the ElevenLabs studio with some oversight but getting it to perfection would probably burn a lot of credits
@grips have you tried some custom workflow with local models?
@matrix not yet, but I doubt my hardware could pull it off at a reasonable speed
Follow
@grips LLM for translation definitely not, but voice models are pretty small
https://github.com/ysharma3501/LuxTTS
> DeepL API pricing includes a free plan that allows up to 500,000 characters per month, suitable for small projects. For more extensive use, the Pro plan starts at a base fee of $5.49 per month plus $25.00 for every additional 1,000,000 characters used.
building the full pipeline for audio could be a bit difficult, but maybe some open solutions already exist