announcement: local speech synthesis. bw LoRAs
Added 2025-07-22 13:19:32 +0000 UTCbringing u style LoRAs and another kick-ass announcement:
upd: the update will be released on sunday
the game will have local speech synthesis from any voice reference file (.wav at least 6 seconds), using the coqui xtts-2 model.
quality’s above average, a bit below openai, runs fast and free, and voice cloning is super easy to use. compiling the binaries for this is a real pain, but I hope to have something for u by friday sunday •ᴗ•
now for the LoRAs:
style b&w manga xl manga
comic style hentai comic random generator
there’ll also be a bugfix, generation speed via sd will be improved, and support for flux models for images will be added - on my laptop I couldn’t comfortably test flux, u need at least 16gb of vram, and it’s much slower than sdxl models
Comments
You can run flux with nunchaku on 8GB VRAM or less, it may be worth it for the better control over scene composition.
Anarchitect
2025-07-23 20:56:52 +0000 UTCnice LoRa , i like this b&w style too !
Youz
2025-07-23 16:06:03 +0000 UTC
