Murmur Transcribes Long Audio 2.5x Faster Than Single-Pass, Beats WhisperX
bariskasikci · x · 2026-11-20
Murmur, from UW's SyFi lab, tackles the long-context ASR trade-off: single-pass decoding is accurate but slow, while chunking is fast but loses speaker and timestamp tracking. Murmur transcribes long audio 2.5x faster than single-pass decoding and beats WhisperX on every benchmark the team tested.
The project was led by undergrad researcher Ruby Lee, who is applying to PhD programs this cycle.
More from Research
- seq2fun assumes fixed cell state; transcriptome-only virtual cell models ignore DNA regulation — anshulkundaje · 2026-11-20
- All-star research group releases timely review of rapidly accelerating AI science field — anshulkundaje · 2026-11-20
- New paper with Yann LeCun asks: can a JEPA learn a latent map of the world? — mengyer · 2026-11-20
- AI Proves Optimal Solutions to MIPLIB Open Problems Unsolved for 9 Years — IgorCarron · 2026-11-20
- Yao Shunyu's private talk notes leak: small-model hierarchies, RSI, small-experiment traps — yangyi · 2026-11-20
- Salesforce releases BLIP-Live: SOTA proactive streaming video understanding with PAUC-F metric — Salesforce · 2026-11-20