smallest_ai's Pulse tops Voice Arena diarization + ASR track with 24.4% DER vs 40.7%
rohanpaul_ai · x · 2026-10-07
smallestai's Pulse ranks #1 on Voice Arena's speaker-attributed ASR track at 24.4% DER (0 ms collar), 1.7x better than runner-up ElevenLabs Scribe v2 at 40.7%. Most of the gap is missed speech: Pulse misses only 4.4% while others exceed 30%. The benchmark scores far-field room audio against labels from per-speaker mics — realistic inputs with accurate ground truth, a rare combination.
More from Models
- you.com search enters RL loop to train Nemotron, lifting BrowseComp to 45.45% — PolarBearby · 2026-10-07
- Student finds Gemini beats ChatGPT for studying: unlimited uploads, interactive diagrams — CommonExplanation711 · 2026-10-07
- OpenAI publishes 722 math papers averaging 3 hours of compute, claiming 90 of top 500 open problems — Latent Space · 2026-10-07
- 'AI will never be good at writing' — one holdout line that's lasted longest — wordgrammer · 2026-10-07
- Ex-Meta ML engineer: local AI models have caught up faster than anyone realizes — Scobleizer · 2026-10-07
- audio.cpp (100+ models, Docker) vs VoiceStudio (27 models): local audio tools compared — pilkyton · 2026-10-07