Everything in this is self-hosted except for the RSS feed and the outgoing emails.
- TTS is using a API version of MeloTTS Docker: GitHub - timhagel/MeloTTS-Docker-API-Server: A docker image to access MeloTTS through API calls
- The AI is running on LM Studio with GPTOSS 20b model
- List of known articles is the in-built Data table because I’m lazy and it works (though it would be better to use the URL rather than the title)
- AI uses a RTX 4070 Super GPU
- TTS is CPU-only (on the same system as the GPU), uses about 15GB of ram
- API call to MeloTTS is batched per 1 item because my system can’t handle too many concurrent requests (so rather 1 call for the full article than concurrent ones split up in chuks)
Tried to merge the output WAV files but haven’t gotten that yet.