[Feature] :Add Voice Input/Output support to AI Agent ChatPanel

[key88cb]

Problem

The current VibeApps desktop environment only supports text-based interaction via the AI Agent ChatPanel. For a system aiming to provide a complete and immersive "desktop experience," lacking voice wake-up and command capabilities (e.g., "Play some jazz," "Open my diary") limits the natural interaction flow and potential accessibility for users.

Proposed Solution

I propose integrating the Web Speech API directly into the existing ChatPanel to enable voice interactions:

1. Speech-to-Text (Voice Input)
   Add a microphone button next to the chat input field in apps/webuiapps/src/components/ChatPanel/index.tsx.
   Utilize the browser's native SpeechRecognition API (handling webkitSpeechRecognition for cross-browser support).
   Once the user speaks, transcribe the voice directly into text and send it as a standard message payload to the AI Agent.
2. Text-to-Speech (Voice Output) [Optional]
   Use the speechSynthesis API to read aloud the Agent's text responses, creating a more realistic "virtual assistant" experience without relying on external TTS services.

Alternatives Considered

Integrating third-party commercial APIs (like OpenAI Whisper or other cloud ASR/TTS services) was considered. However, prioritizing the browser's native Web Speech API aligns perfectly with OpenRoom's current architectural philosophy of being "zero-backend" and working "out-of-the-box" entirely within the client environment.

Additional Context

This would likely require adding a permission prompt flow for microphone access upon the first click.
A visual indicator (like a pulsing animation or waveform) during active listening would greatly improve UX.

[key88cb]

well sry to interrupt, but I m curious about MiniMax's attitude for the AIRI x Iru project. Will U develop it to be more like AIRI, or focus primarily on the Agentic UI?

[DanteAlighieri13210914]

Thank you so much for the suggestion! We’ll definitely consider it.
As for the AIRI x Iru project, we hadn’t heard of it before. OpenRoom is just a side project exploring agent scaffolds for entertainment scenarios. But we’ll look into that project later. Glad to see other explorations in this space!