53 of 53
| Job | Licence | Commercial | TurnBench | eot-bench | |||||
|---|---|---|---|---|---|---|---|---|---|
Backchannel Cue 32MKolja Beigel | Backchannel | MIT | Yes | 32M | 1 | 2026-11 | — | — | — |
Soniox stt-rt-v6Soniox | Streaming ASR | Proprietary | Yes | — | 60+ | 2026-11 | — | — | 647 ms |
ai-coustics Quail Voice Focus 2.3ai-coustics | Noise and echo | Proprietary | Yes | — | — | 2026-11 | — | — | — |
tygerNineNineSix | Turn detection | OpenMDW-1.1 | Yes | 652M | 1 | 2026-10 | — | — | — |
turn-1-minip99lab | Turn detection | Apache-2.0 | Yes | 5.3M | 1 | 2026-10 | 0.975 | 595 ms | — |
VoxTurnUniversity of Melbourne | Turn detection | CC-BY-NC-4.0 | No | 800M | 14 | 2026-10 | 0.924 | 557 ms | — |
MaAIKyoto University | Turn detection | Custom terms | Restricted | — | 3 | 2026-10 | — | — | — |
Pipecat backchannel detectorPipecat | Interruption | BSD-2-Clause | Yes | 734K | 1 | 2026-10 | — | — | — |
| Interruption | CC-BY-NC-4.0 | No | — | 1 | 2026-10 | — | — | — | |
Voxtral Mini Transcribe Realtime 3Mistral AI | Streaming ASR | Proprietary | Yes | — | — | 2026-10 | — | — | — |
VoxLockArbaz Siddiqui | Noise and echo | Apache-2.0 | Yes | 27M | 1 | 2026-10 | — | — | — |
Alkove FocusAlkove | Noise and echo | Apache-2.0 | Yes | — | — | 2026-10 | — | — | — |
Echo OmniZero Runtime | Turn detection | Proprietary | Unclear | — | 27 | 2026-09 | 0.839 | 676 ms | — |
Echo Omni DuplexZero Runtime | Turn detection | Proprietary | Unclear | — | 1 | 2026-09 | 0.970 | 768 ms | — |
 AssemblyAI Universal-3.6 Pro RealtimeAssemblyAI | Streaming ASR | Proprietary | Yes | — | 32 | 2026-09 | — | — | 1049 ms |
 Cartesia Ink-2Cartesia | Streaming ASR | Proprietary | Yes | — | 5 | 2026-09 | — | — | 1056 ms |
Audio8 ASR InfiniteEdge0 | Streaming ASR | Custom terms | Unclear | 4.1B | 2 | 2026-09 | — | — | — |
X2-Turn-4BX Square Robot | Streaming ASR | Permissive | Yes | 4B | 2 | 2026-08 | — | — | — |
| Noise and echo | Proprietary | Yes | — | — | 2026-08 | — | — | — | |
 LiveKit Turn Detector v1LiveKit | Turn detection | Custom terms | Yes | — | 14 | 2026-06 | — | — | 543 ms |
DualTurn EndpointingAnyreach AI | Turn detection | Apache-2.0 | Yes | — | 1 | 2026-06 | — | — | — |
Mimi EndpointerSathvik Udupa et al. | Turn detection | CC-BY-4.0 | Yes | — | 1 | 2026-06 | 0.782 | 645 ms | — |
| Turn detection | Proprietary | Yes | 9M | 12 | 2026-05 | — | — | — | |
TurnSenseBairong | Turn detection | Custom terms | Unclear | 47M | 2 | 2026-05 | — | — | — |
| Interruption | Proprietary | Yes | 6M | 1 | 2026-05 | — | — | — | |
| Streaming ASR | Proprietary | Yes | — | 38+ | 2026-05 | — | — | — | |
 Deepgram FluxDeepgram | Streaming ASR | Proprietary | Yes | — | 10 | 2026-04 | — | — | 1151 ms |
| Interruption | Proprietary | Yes | — | — | 2026-03 | — | — | — | |
FireRedVADFireRedTeam (Xiaohongshu) | Voice activity | Apache-2.0 | Yes | — | 100+ | 2026-02 | — | — | — |
Smart Turn v3.2Pipecat | Turn detection | BSD-2-Clause | Yes | 8M | 23 | 2026-01 | 0.752 | 1017 ms | 1051 ms |
Picovoice CobraPicovoice | Voice activity | Proprietary | Yes | — | — | 2025-12 | — | — | — |
Picovoice KoalaPicovoice | Noise and echo | Proprietary | Yes | — | — | 2025-12 | — | — | — |
| LiveKit text turn detectorLiveKit | Turn detection | Custom terms | Yes | 500M | 14 | 2025-11 | — | — | — |
Silero VADSilero | Voice activity | MIT | Yes | — | 6000+ | 2025-11 | — | — | — |
Vogent-Turn-80MVogent | Turn detection | Custom terms | Restricted | 79M | 1 | 2025-10 | — | — | — |
| Streaming ASR | Custom terms | Yes | 120M | 1 | 2025-10 | — | — | — | |
Namo Turn Detector v1VideoSDK | Turn detection | Apache-2.0 | Yes | — | 23 | 2025-09 | — | — | — |
 UltraVADUltravox | Turn detection | Custom terms | Unclear | — | 26 | 2025-09 | — | — | 899 ms |
Easy TurnASLP Lab, Northwestern Polytechnical University | Turn detection | Apache-2.0 | Yes | — | 2 | 2025-09 | — | — | — |
FireRedChat turn detectorFireRedTeam (Xiaohongshu) | Turn detection | Apache-2.0 | Yes | — | 2 | 2025-09 | — | — | — |
FireRedChat personalized VADFireRedTeam (Xiaohongshu) | Voice activity | Apache-2.0 | Yes | — | 2 | 2025-09 | — | — | — |
TEN VADTEN Framework (Agora) | Voice activity | Custom terms | Restricted | — | — | 2025-07 | — | — | — |
 Kyutai STT 1BKyutai | Streaming ASR | Permissive | Yes | 989M | 2 | 2025-06 | 0.773 | 1007 ms | — |
| Voice activity | Custom terms | Yes | 92K | 6 | 2025-05 | — | — | — | |
TEN Turn DetectionTEN Framework (Agora) | Turn detection | Custom terms | Restricted | 7.6B | 2 | 2025-04 | — | — | — |
ESPnet turn-taking predictionESPnet (CMU) | Turn detection | CC-BY-4.0 | Yes | — | 1 | 2025-03 | 0.826 | 862 ms | — |
| Turn detection | Proprietary | Yes | — | — | 2025-03 | 0.303 | 793 ms | 1143 ms | |
WebRTC AEC3WebRTC project | Noise and echo | BSD-3-Clause | Yes | — | — | 2025-01 | — | — | — |
RNNoiseXiph.Org | Noise and echo | BSD-3-Clause | Yes | — | — | 2024-04 | — | — | — |
DeepFilterNet3Hendrik Schröter | Noise and echo | Permissive | Yes | — | — | 2023-08 | — | — | — |
SpeexDSPXiph.Org | Noise and echo | Permissive | Yes | — | — | 2022-06 | — | — | — |
Voice Activity Projection (VAP)KTH (Ekstedt and Skantze) | Turn detection | Permissive | Yes | — | 1 | 2022-05 | 0.845 | 368 ms | 1131 ms |
WebRTC VADWebRTC project | Voice activity | Permissive | Yes | — | — | 2017-01 | — | — | — |
