# VoxLock: Target-Speaker Extraction for Voice Agents Without an Enrollment Step

**Authors:** Arbaz Siddiqui  
**Published:** 29 Oct 2026 · **Upvotes:** 4  
**Links:** [arXiv](https://arxiv.org/abs/2610.40727) · [PDF](https://arxiv.org/pdf/2610.40727) · [This page](https://omen-bit.github.io/ai-paper-daily/paper/2610.40727.html)

## Key numbers

- **4.95%**: On Krisp's 265-recording benchmark it reaches 4.95% WER against 23.21% unprocessed and lets 9 background words…
- **23.21%**: …Krisp's 265-recording benchmark it reaches 4.95% WER against 23.21% unprocessed and lets 9 background words into…
- **42.95% → 10.63%**: it cuts WER from 42.95% to 10.63%

## Abstract

A voice agent's speech recognizer transcribes everyone near the microphone, including people other than the caller. I fine-tuned a public causal target-speaker extraction model (REAL-TSE, 27M parameters) on one GPU to work from a voice sample cut from the call itself, and released it as VoxLock. I score it by what a recognizer writes on real recordings. On Krisp's 265-recording benchmark it reaches 4.95% WER against 23.21% unprocessed and lets 9 background words into the transcript over 85 minutes of background-only speech. Its WER is not significantly different from Krisp's best filters and is significantly lower than the other commercial and open systems I tested. On 400 AMI meeting windows built and frozen before scoring, it cuts WER from 42.95% to 10.63%. The 8-bit model runs a live call on 77% of one CPU thread with 24 ms of algorithmic latency. Model, code and per-file scores are open.

---

Notes written by AI from the paper, numbers copied from the abstract. Check the paper before you rely on them. From AI Papers Daily.