wllama.cpp omni demo

Omni: voice chat, text to speech, speech to text

Step 1: Load a model

Model:

Or pick local GGUF files (select both the model and the mmproj file):

Voice chat (speech to speech)

Hands-free: once started, just talk. When you stop talking, the model answers with text and voice, then listens again. Headphones work best.

Realtime speech to text

Text to speech

Text:

Voice:

Voice to clone: none
A few seconds of clear speech, recommended (without it the voice may sound wrong). WAV, MP3 or FLAC, other formats must be converted first.

Language:

Transcribe an audio file

none

Language: