Feature Request: Add Voice Input/Output Button to Chat Input Box

[gfprimroseking-maker]

Feature Request: Add Voice Input/Output Button to Chat Input Box

🎤 Feature Description

Add a microphone button 🎤 to the chat input box in ClawX, enabling users to speak their messages directly and optionally have AI responses read aloud via TTS (Text-to-Speech).

💡 Use Case

I use ClawX as a novel-writing assistant and other creative tasks. Sometimes I don't want to type — I want to talk to the AI naturally, like having a conversation. This would make ClawX much more accessible and comfortable for:

• Hands-free operation — cooking, driving, walking while chatting
• Accessibility — users who have difficulty typing (elderly, motor impairments)
• Creative brainstorming — speaking ideas is often more natural than typing them
• Longer messages — speaking is faster than typing for longer thoughts
• Non-English input — some users are more comfortable speaking than typing in certain languages

🖼️ Proposed UI Design

Chat Input Box

• 🎤 Microphone button — click to start/stop voice recording
• Clicking activates Windows Speech Recognition API (or Web Speech API)
• The button changes state while recording (e.g., pulse animation, red dot)
• Transcribed text appears in the input box for review before sending
• Optional: "Send on complete" toggle for hands-free sending

Output (Optional Enhancement)

• 🔈 Speaker button next to AI responses — click to hear the response read aloud
• Uses the existing TTS capability (ElevenLabs, OpenAI, or system TTS)
• Optional auto-read setting for accessibility mode

🔧 Technical Approach

ClawX is an Electron-based desktop app built on OpenClaw. Here are possible implementation paths:

Option A: Web Speech API (Browser-based, simplest)

• Uses the browser's built-in SpeechRecognition API
• Zero dependencies, works out of the box
• Limitation: Requires Chromium 25+ (already satisfied in Electron)
• Limitation: Chrome only, needs online connection

Option B: Whisper.cpp (Local, offline)

• Integrate whisper.cpp as a native Node.js module
• Full offline STT, no API costs
• Higher initial setup complexity but best privacy

Option C: Cloud API (OpenAI Whisper API / Azure Speech)

• Best accuracy, supports many languages
• Requires API key configuration in settings
• Can be a configurable option in ClawX settings

For TTS Output:

• OpenClaw already has tts tool support — just needs a UI button
• Could use system TTS (Windows SAPI, macOS NSSpeechSynthesizer) or cloud TTS

✅ Benefits

1. Competitive advantage — Most AI desktop assistants don't have native voice input
2. User retention — Voice interaction is more engaging and natural
3. Accessibility compliance — Helps meet WCAG guidelines
4. Low implementation cost — Web Speech API is free and works immediately

📋 Configuration Ideas (Settings Panel)

• Enable Voice Input
  • STT Engine: Windows Speech API / Web Speech API / OpenAI Whisper
  • Language: Chinese (Mandarin) / English / Auto-detect
  • Auto-send on complete: On/Off
• Enable Voice Output (TTS)
  • TTS Engine: System TTS / ElevenLabs / OpenAI TTS
  • Voice: Select voice
  • Read responses automatically: On/Off

🙏 Request

I'd love to see this feature added to ClawX. If the core team is busy, I'm happy to contribute or collaborate on this. Voice interaction would make ClawX feel much more like a real assistant and less than a text box.

Thanks for building ClawX! It's an amazing tool. 🐾

[richardchen874-sys]

Voice input/output is a good feature to make configurable because STT, TTS, and chat reasoning all have different cost/privacy tradeoffs. A local-first default with optional cloud models sounds like the right shape.

For the cloud path, I’d separate the pipeline:

• local STT where privacy or cost matters
• cheap model for command cleanup / short intent extraction
• stronger model only for complex reasoning
• per-session token/cost display so voice loops stay predictable
• provider profiles for OpenAI-compatible alternatives

I'm building a private-beta OpenAI-compatible API aggregation layer focused on Chinese models such as DeepSeek / Qwen / GLM-style providers, using official-provider keys. If you want to benchmark lower-cost chat models for the assistant side of ClawX, I’d be happy to share some test credits.

Curious: would users prefer offline voice by default, or a cloud option with better accuracy and visible cost?