AI BriefingLFM3-Audio-2B: LFM3-Audio: A 2B Parameter Model That Listens to 19 Languages
LiquidAI/LFM3-Audio-2B
About the project
LFM3-Audio-2B, released by LiquidAI, is an automatic speech recognition (ASR) model that converts audio input into text. It is a small model with a total of 2 billion (2B) parameters, designed to run on edge devices or in low-spec environments.
It supports a total of 19 languages, including English, Korean, Japanese, Chinese, Arabic, German, and French. It recorded an average WER (Word Error Rate) of 6.23% on the hf-audio/open-asr-leaderboard, ranking 13th on that benchmark.
It supports the Transformers library and the safetensors format, and carries the any-to-any pipeline tag. Specialized for conversational and multilingual processing, it can be used not only for simple transcription tasks but also for building voice-based dialogue systems.
It is suitable for developers who need to process voice data in a local environment instead of a cloud server. Being lightweight at the 2B scale, it lowers hardware requirements compared to existing large ASR models while maintaining recognition performance for major languages.
LiquidAI/LFM3-Audio-2B
The original page has no description.
any-to-any
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.