Text To Speech

tortoise TTS is too slow and making good AI models takes time websites like elevenlabs and fakeyou are too expensive for mass production of audio. You can get away with shorter stuff for free, however. for TTS i still use festival scripts piped through libreoffice writer TTS plugin. I can get Linux or Windows to generate an 10 hour audiobook from a pile of text ripped out of a PDF in about a half an hour. Not bad. problem is, it sounds like soykaf.

>>12615 This is basically what I do. I've gotten used to the voice. But I don't get what you are using libreoffice writer tts for? Are you using it to handle some rare characters better? I just use text2wave on a regular text file, and I definitely generate a 10 hour audiobook way faster than a half hour on a half-dead computer that rattles if I run it too hot. I'd be surprised if it took even five minutes.

>>12620 lain probably just uses it because it's free

>making good AI models takes time Yes, but you should invest in it if you're serious. There's tons of public domain stuff you can train it on if you care about copyright protection.

have you tried these? https://github.com/Burgstall-labs/ComfyUI-BS_Kokoro-onnx https://www.zyphra.com/post/beta-release-of-zonos-v0-1 https://github.com/Zyphra/Zonos/ https://github.com/SparkAudio/Spark- https://github.com/canopyai/Orpheus-TTS