Real-time Speech-to-Speech Interaction Latency Evaluation
0.8User-Visible Response LatencySesame web app
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Sesame web appInteraction=speech-to-speech, Comparison boundary=Vendor benchmark; speech-only.2026.06 | 0.8 | — | — | — | — | — | |
| Hume EVI 3Interaction=speech-to-speech, Comparison boundary=Vendor benchmark; no visual output stream.2026.06 | 0.9 | 300 | — | — | — | — | |
| Doubao Realtime VoiceInteraction=speech-to-speech, Comparison boundary=Official speech-only product numbers; no visual agent output.2026.06 | 1 | 700 | — | — | — | — | |
| Gemini Live APIInteraction=speech-to-speech, Comparison boundary=Vendor benchmark; not an official model breakdown.2026.06 | 1.2 | — | — | — | — | — | |
| MoshiInteraction=speech-to-speech, Comparison boundary=Native full-duplex speech model; no visual agent.2026.06 | — | 160 | — | — | — | — | |
| SeeduplexInteraction=speech-to-speech, Comparison boundary=Relative production improvement; speech-only.2026.06 | — | — | — | 250 | 300 | — |