Liquid AI’s audio model in one piece. The same model listens, decides what to say and speaks, so there is no pipeline to choose. The smallest way to start, on a Mac or on Linux.
npx boubou@beta run LFM3-Audio-2B
Talk on this Mac. It needs Node.js 18 or later; Get started walks you through.
Talk on this Linux PC. It needs Node.js 18 or later; Get started walks you through.
npx boubou@beta serve LFM3-Audio-2B
Talk from a phone on the same Wi-Fi. The model still runs on this Mac. Phones and robots walks you through.
Talk from a phone on the same Wi-Fi. The model still runs on this PC. Phones and robots walks you through.
This version runs on Apple Silicon Macs only. On Linux, choose LFM3-Audio-2B.
Apple Silicon Mac
x86-64 Linux PC
Memory needs are estimates. Boubou checks the machine before it downloads anything, and smaller machines still need testing. On arm64 Linux, build Boubou from source.
Everything the model can speak, as its own card or files list it. Black marks what Boubou uses.
Languages
Boubou’s voice chat with this model is English. The other languages are listed on the model card.
Source: LiquidAI/LFM3-Audio-2B-GGUF
Voices
Press a voice to hear it. Each clip is the pinned model reading one sentence in that voice’s language.
Speech synthesis voices
Boubou’s conversation prompt does not choose a voice, so it is not fixed. llama-tts uses female unless told otherwise.
Source: LiquidAI/LFM3-Audio-2B-GGUF
Boubou audio helper
echo cancellationLFM3-Audio-2B
hears, answers and speaks · llama.cpp b11960Boubou owns the microphone and speakers. Apple voice processing on a Mac, or WebRTC AEC3 on Linux, cancels echo, so you can talk over a reply.
LFM Open License. Read the license. Weights keep their own licenses; check them before commercial use.
Every file must match its SHA-256 before Boubou uses it.
| File | What it is | Size | SHA-256 |
|---|---|---|---|
LFM3-Audio-2B-Q4_K_M.gguf | Model, Q4_K_M | 1.10 GB | 9a773289…5be2 |
mmproj-LFM3-Audio-2B-Q8_0.gguf | Audio projector, Q8_0 | 304 MB | 2b6648b1…e365 |