p99lab turn-1, an end-of-turn detection model
Copyright 2026 p99lab (https://p99lab.com)

This product includes turn-1, an end-of-turn detection model developed by p99lab
(https://huggingface.co/p99lab/turn-1). Redistributed copies, and models or products that include or are
derived from it, must retain this notice, as section 4(d) of the Apache License, Version 2.0 requires.
The names "p99lab" and "turn-1" may not be used to name or endorse derived models or products without
permission (section 6).

The code in this repository and the weight files "turn-1-head.safetensors" and "turn-1-lora.safetensors" are released
under the Apache License, Version 2.0 (see LICENSE).

Third-party models and code
turn-1 runs on the two models below. They are downloaded from their own repositories at run time, are not
redistributed here, and stay under their own licence. Their authors do not endorse p99lab or this model.

Qwen3-ASR-1.7B, Alibaba Cloud Qwen team. Apache License 2.0. https://huggingface.co/Qwen/Qwen3-ASR-1.7B
Qwen3-ASR-0.6B, Alibaba Cloud Qwen team. Apache License 2.0. https://huggingface.co/Qwen/Qwen3-ASR-0.6B
Qwen3-ASR code: https://github.com/QwenLM/Qwen3-ASR and the PyPI package qwen-asr 0.0.6. Apache License 2.0.
  p99turn1/turn1.py contains a batched re-implementation of the forward pass of the Qwen3-ASR audio encoder,
  derived from qwen_asr/core/transformers_backend/modeling_qwen3_asr.py. Changes: batching over windows of at
  most one attention block; hidden states returned at three depths.

Third-party data credits
The datasets below are licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), which asks for credit.
The model was trained in part on them. Their audio was processed for training and is not redistributed here.
Their authors do not endorse p99lab or this model.

HarperValleyBank. C. Wu, M. Nafziger, A. Scodary, A. Maas, "HarperValleyBank: A Domain-Specific Spoken Dialog
  Corpus", arXiv:2010.13929. https://github.com/cricketclub/gridspace-stanford-harper-valley
LibriTTS-R. Y. Koizumi et al., "LibriTTS-R: A Restored Multi-Speaker Text-to-Speech Corpus", Interspeech 2023;
  H. Zen et al., "LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech", Interspeech 2019.
  https://www.openslr.org/141/
CSTR VCTK Corpus (version 0.92). J. Yamagishi, C. Veaux, K. MacDonald, University of Edinburgh, CSTR, 2019.
  https://doi.org/10.7488/ds/2645
EVI. G. Spithourakis et al., "EVI: Multilingual Spoken Dialogue Tasks and Dataset for Knowledge-Based Enrolment,
  Verification, and Identification", Findings of NAACL 2022 (arXiv:2204.13496). PolyAI. https://github.com/PolyAI-LDN/evi-paper
ICSI Meeting Corpus. A. Janin et al., "The ICSI Meeting Corpus", ICASSP 2003. https://groups.inf.ed.ac.uk/ami/icsi/

Evaluation (never trained or tuned on)
eot-bench (code), LiveKit. Apache-2.0. https://github.com/livekit/eot-bench
livekit/eot-bench-data, LiveKit. CC BY 4.0 per its dataset card. https://huggingface.co/datasets/livekit/eot-bench-data