openclaw - 💡(How to fix) Fix Feature Request: Native local Whisper integration for audio transcription [1 participants]

TL;DR

Implement a local OpenAI Whisper audio transcription integration, similar to the existing Groq integration, to enable automatic transcription without requiring external API calls or exec approvals.

Guidance

• Investigate modifying the existing Groq audio transcription code to call the local Whisper binary instead of making an API call, leveraging the similar integration pattern.
• Ensure that the Whisper CLI and ffmpeg are properly installed and configured on the system, as they are required for audio format conversion and transcription.
• Explore implementing allowlisting for Whisper commands to prevent variable file paths in arguments from causing issues, potentially by using a fixed directory for temporary audio files.
• Review the local Whisper skill implementation for insights into integrating Whisper with the OpenClaw setup.

Example

No explicit code example can be provided without more context, but the Groq audio transcription documentation and local Whisper skill code can serve as references for implementing the local Whisper integration.

Notes

The success of this integration depends on the compatibility of the Whisper CLI with the OpenClaw setup and the ability to properly handle audio file paths and conversions.

Recommendation

Apply a workaround by modifying the existing Groq integration to use the local Whisper binary, as this approach leverages existing code and minimizes the introduction of new dependencies or external API calls.