Enhance Text-to-Speech (TTS) Integration with Audio File Management and Playback

[johnbenac]

This update introduces a comprehensive enhancement to the Text-to-Speech (TTS) feature, focusing on improving audio file management, storage, and in-message playback capabilities. Key changes include the creation of a dedicated endpoint for audio uploads, utility function enhancements for file handling, and user interface adjustments to facilitate audio control within messages, and partial migration to event-driven automatic audio generation. Here's a breakdown of the modifications:

New Audio Upload Endpoint (/uploadaudio): We've implemented a server-side endpoint to handle the upload and storage of generated TTS audio files. This endpoint accepts base64 encoded audio data, processes it, and saves it to a designated directory on the server.

Utility Function for File Saving (saveAudioAsFile): A new utility function has been added to convert audio data from base64 format to a file, saving it within the server's file system. This facilitates easy management and retrieval of TTS audio files.

Enhanced appendMediaToMessage Function: Modifications to this function allow for the dynamic insertion of audio elements within chat messages. Users can now directly play, pause, and download TTS-generated audio without leaving the chat interface.

Frontend Adjustments in index.html: Minor updates to the HTML structure support the display and control of embedded audio files within messages, ensuring a seamless user experience.

Script Updates (tts/index.js): Key script enhancements link the TTS feature with the new audio management capabilities. These changes ensure that TTS audio is correctly generated, saved, and associated with the corresponding message, complete with UI updates for immediate playback.

_Right now, the user message generation is broken. It can be manually generated, but not automatically generated, as the user entering a message does not emit an event, and the logic isn't smart enough to intelligently hunt through recent state for unrendered user audio

Together, these changes significantly improve the TTS feature by not only enriching the user's interactive experience with instant audio feedback but also by streamlining the backend management of TTS audio files. This update lays the groundwork for further enhancements in audio-based communication within SillyTavern.

[johnbenac]

Per our conversation in discord:(https://discord.com/channels/1100685673633153084/1100820587586273343/1218600978119393300)

I've improved the trigger by using better message, so the autogeneration now works on user messages. I've tested it in groups and with streaming on and off and user messages being narrated and not, and it all works as it should.