Blog
Release notes, platform updates, and product news from the VoiceThere team.
Tags: voiceClear filters
October 21, 2026 · release
VoiceThere 0.3.0
Speech on dedicated servers with faster, more accurate replies under load, language switching mid-call, calls that survive network drops, deploy history and a redesigned dashboard.
@voicethere/agent 0.10.0 · @voicethere/cli 0.16.9 · @voicethere/client 0.11.9
October 20, 2026 · release
One request shape for every speech provider
The built-in speech is now called voicethere, and setVoiceLanguage takes the same request for VoiceThere and every cloud vendor, with error messages that say what to change.
@voicethere/agent 0.10.0 · @voicethere/cli 0.16.9
October 20, 2026 · platform
Speech latency numbers, published
Median and p90 latency for VoiceThere speech and five cloud speech vendors, with the method and the commands to repeat it.
node-webrtc-rust 0.12.1 · @voicethere/client
October 19, 2026 · platform
Load testing VoiceThere: 45 voice conversations, measured
Detailed results of our voice load tests: 45 conversations at once, capacity per pair of speech servers, and where the wait between the end of a sentence and the agent's answer comes from.
October 17, 2026 · release
Fish Audio Drama 3 for voice agents
Pick Fish Audio for TTS with Drama 3, for STT with transcribe-1, or both, using your own Fish key.
October 16, 2026 · platform
Voice replies start sooner after a period
Voice agents now capitalize the first letter after a sentence end in the text sent to the voice engine, so the engine splits sentences there. A project setting turns it off.
October 13, 2026 · platform
Language switches get a second try
A slow language start is announced to the caller, retried once, and a language that still cannot start is reported instead of being misheard.
October 10, 2026 · release
See which voice used your characters
The project Usage page now breaks speech usage down by voice, with characters, credits and sessions per voice.
October 8, 2026 · platform
Project metrics: see how your agent performs
A new Metrics area with Overview, Sessions, Voice and Agent tabs, plus a separate Usage page for credits.
October 8, 2026 · release
Short network drops no longer end a voice conversation
@voicethere/client 0.11.6 rides out short network drops, the reconnect window is a project setting, and replies wait for a busy speech server instead of dropping the call.
@voicethere/client 0.11.6
October 6, 2026 · platform
Repeated replies now play from memory, free of charge
Voice agents replay repeated TTS replies from memory, those replays are not billed, and a project setting lets you turn the cache off.
October 5, 2026 · release
Language switches without interrupting the caller
Automatic language switches now wait for the caller to finish, say a short ready message after a slow switch, and need VoiceThere speech on both sides.
September 30, 2026 · platform
New guide: pauses and pronunciation for VoiceThere voices
We added documentation on shaping how VoiceThere voices speak your agent's text.
September 29, 2026 · release
Mid-conversation language switch for voice agents
setVoiceLanguage keeps WebRTC up while STT/TTS follow the caller; optional LID auto-switch and per-language presets are off by default.
@voicethere/agent
September 28, 2026 · release
Expanded cloud STT and TTS vendors
Configure Deepgram TTS, ElevenLabs Scribe, Groq, Azure Speech, and AWS alongside existing vendors, with per-vendor setup docs and project secrets.
September 10, 2026 · platform
Tune spoken language identification from the dashboard
Enable or disable Whisper LID and set minSpeechMs (1000–5000 ms, default 2500) per project.
September 7, 2026 · release
Spoken language identification for voice agents
Detect the caller’s spoken language and handle user_language events to route prompts and TTS.
@voicethere/agent
September 1, 2026 · release
Inbound RNNoise on by default for voice runners
Voice and Voice+Data projects now enable inbound noise suppression by default; turn it off per project when CPU is tight.
August 27, 2026 · release
Live showcase demos on the marketing site
Try voice and multiplayer examples from the home page and /showcase without signing up first.