LavaSR: A 50MB audio enhancer running 5,000x faster than real-time
LavaSR just showed up as an ultralight audio enhancement tool that cleans up noisy speech almost instantly. Weighing just 50MB, it runs up to 5,000x real-time on GPUs and 60x on CPUs, making it small and fast enough to run locally on mobile phones. Why it matters: It makes real-time, zero-lag voice cleanup accessible without needing heavy cloud compute.
LavaSR is an ultralight 50MB audio enhancer designed to strip out background noise and restore muffled speech with minimal compute.
What it does and how to run it
Its biggest draw is sheer speed. LavaSR clocks in at 5,000x real-time on a GPU and still hits 60x real-time on a standard CPU, making it lightweight enough to run entirely on a mobile phone.
Official demos show noticeably cleaner vocals with much stronger presence after processing.
You can test it right now via a free Hugging Face Space. Just upload a noisy clip, leave the settings at default, and click Enhance for near-instant results. The creators also provide a Google Colab notebook and full local deployment instructions.
Why it matters
If you edit podcasts or build apps that need fast, local voice cleanup, LavaSR is definitely worth a test run.
While heavy generative audio models demand multi-gigabyte downloads and pricey cloud GPUs, LavaSR proves that voice enhancement can be tiny, ultra-fast, and cheap to run right on everyday edge devices.
Sources (3)
- github.com(opens in a new tab)Originalgithub.com
- huggingface.co(opens in a new tab)Reporthuggingface.co
- huggingface.co(opens in a new tab)Reporthuggingface.co