LFM2.5-Audio-1.5B

Integrated English Voices not listed Mac · Linux

Liquid AI’s earlier audio model, served by the Liquid Audio runner. It is smaller, but interrupting a reply takes longer because Boubou has to restart its server. It is a developer choice and is not offered in the menu.

$ npx boubou@beta run LFM2.5-Audio-1.5B

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 LFM2.5-Audio-1.5B

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 LFM2.5-Audio-1.5B.

Download: 1.27 GB1.27 GB
Memory, estimated: —
Speaks: English
Voice: Model default

Recommended device

Apple Silicon Mac

No memory estimate for this version yet
macOS 13.3 or later
Echo cancellation is built in, so headphones are optional
The voice runs on the Liquid Audio runner

x86-64 Linux PC

No memory estimate for this version yet
glibc 2.38 or later, such as Ubuntu 24.04
No GPU needed: the model runs on the CPU
Echo cancellation is built in, so headphones are optional

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.

Voices and languages

Everything the model can speak, as its own card or files list it. Black marks what Boubou uses.

Boubou uses
Available in the model

Languages

English

The model card lists English.

Source: LiquidAI/LFM2.5-Audio-1.5B-GGUF

Voices

Voices not listed

The model card does not name its voices.

Source: LiquidAI/LFM2.5-Audio-1.5B-GGUF

How it works

You speak

AUDIO

Boubou audio helper

echo cancellation

ONE MODEL

LFM2.5-Audio-1.5B

hears, answers and speaks · the Liquid Audio runner

It answers

Boubou 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.

License

LFM Open License. Read the license. Weights keep their own licenses; check them before commercial use.

Sources and files

LiquidAI/LFM2.5-Audio-1.5B-GGUF

Every file must match its SHA-256 before Boubou uses it.

File	What it is	Size	SHA-256
`LFM2.5-Audio-1.5B-Q4_0.gguf`	Model, Q4_0	696 MB	`3583bee8…e0d8`
`mmproj-LFM2.5-Audio-1.5B-Q8_0.gguf`	Audio projector, Q8_0	293 MB	`8f943619…9c67`
`tokenizer-LFM2.5-Audio-1.5B-Q8_0.gguf`	Audio tokenizer, Q8_0	77 MB	`74f7f648…15e3`
`vocoder-LFM2.5-Audio-1.5B-Q8_0.gguf`	Vocoder, Q8_0	206 MB	`3063b8eb…3f23`