Repository navigation
Replies: 1 comment
|
+1, would use this a lot. Ideally one long-running voice conversation on mobile that manages all my threads (different models, in parallel) without opening each one. Some notes:
|
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Summary
Feature proposal for a Hands-Free Voice Orchestrator mode in T3 Code (Desktop & Mobile), combining voice task dispatching across threads with ultra-fast, zero-token-cost neural speech synthesis (using Microsoft Edge Neural TTS or local Web Speech API).
Motivation
When orchestrating multiple concurrent coding agents (Claude Code, Codex, Antigravity, OpenCode), reading long wall-of-text diffs and terminal outputs while switching between keyboard and mice breaks the developer flow.
In advanced agentic environments (such as Antigravity harness and voice-driven agent supervisor setups), having a Voice Orchestrator allows developers to control threads completely hands-free:
Core Components
1. Voice Dispatcher & Multi-Thread Routing
A persistent voice conductor widget (or global hotkey / push-to-talk) that accepts speech input, transcribes it, and routes the command:
2. High-Speed, Zero-Cost Neural Speech Output (Edge-TTS Protocol)
Instead of wasting expensive LLM tokens or paid third-party voice APIs (OpenAI TTS / ElevenLabs), speech can be streamed at practically zero latency and zero marginal cost using the Microsoft Edge Neural Speech protocol:
en-US-AriaNeural,en-US-GuyNeural, plus 100+ languages).--rate=+50%to+100%without audio distortion.3. Progressive Spoken Turn Summaries ("What's happening / What was done")
Instead of reading full code diffs, agents can emit an audio digest on turn completion:
apps/server, fixed the projection leak, test suite passed."rm -rf tmp/cache. Approve or deny?" — and accepts a spoken "Approve".Technical Feasibility & Reference Implementation
We have tested this pattern in production with a lightweight local script (
ozvuchka-tts.mjs/ Edge-TTS wrapper) streaming directly to playback buffers. A typical turn summary takes < 400ms to begin streaming audio.In T3 Code architecture:
orchestration.dispatchCommand(thread.create,session.prompt,permission.respond).apps/desktopaudio player.Would love to hear thoughts from the team and community on integrating a native Voice / Hands-Free mode into T3 Code!
All reactions