Sneak Peek: The Next Big Update for Synthalingua

Hello everyone! As a solo developer, I've been working hard behind the scenes on a massive architectural update for Synthalingua, and I'm incredibly excited to share a preview of what's coming next.

This next update (which will be Beta 1.1.1) is focused on making the application faster, more stable, and packed with powerful new tools. While it's not ready for release just yet, here’s a look at what I've been building.

Major New Features on the Horizon

These are brand-new capabilities designed to solve common frustrations and unlock new possibilities.

1. Crystal-Clear Transcriptions with AI Vocal Isolation

This is a game-changer for anyone working with less-than-perfect audio. A new feature will use an AI model (Demucs) to intelligently separate spoken words from background music, in-game sounds, noise, and other interference. This means you'll be able to get clean, high-quality transcriptions from audio that was previously unusable.

2. Interactive Stream Selection with Audio Preview

Tired of transcribing the wrong audio from a livestream? Soon, when you target a live stream, Synthalingua will show you all available audio tracks. More importantly, it will let you download and listen to a short preview of any track before starting the full transcription. This eliminates the guesswork and ensures you're always transcribing the correct source.

3. More Natural Live Microphone Transcription

I'm improving the way the AI handles live audio. A new context-aware mode will allow the AI to remember what you just said, using it as a reference for what you're about to say next. The result will be far more fluid and accurate live transcriptions with fewer awkwardly cut-off sentences.

4. Advanced Subtitle Generation

The subtitle generator is getting a massive upgrade. A new Model Comparison Mode will let you generate subtitles using every available AI model with a single command, so you can easily compare the output files and choose the best one. I'm also adding support for Word-Level Timestamps to create karaoke-style subtitles with precise timing.

Fixes, Improvements, and Stability Upgrades

Alongside new features, I've focused heavily on improving the core experience and fixing common issues.

• A Modern, Readable Interface: The command-line output has been completely redesigned with clean formatting, colors, and structured boxes to make it much easier to read at a glance.
• Smarter Content Filtering: The application will now be able to detect when the AI gets stuck repeating the same phrase (a common issue called "looping") and will automatically block the repetitive output to keep your transcriptions clean.
• Overhauled Remote Microphone Tool: The remote_microphone.py script has been transformed into a powerful, standalone utility. It will allow you to set up remote audio streaming from one computer to another over your local network, complete with a live audio meter and device testing.
• More Resilient Streaming: If a streaming service temporarily blocks the connection, the application will now intelligently wait and retry instead of crashing. This will make long-running streams far more stable.
• Automatic Memory Cleanup: The tool will now automatically clear the AI model from your computer's memory (RAM/VRAM) after each file is processed, preventing slowdowns when you're working with multiple files.
• Easier Setup: Getting started will be even simpler. You'll no longer need to provide an exact path to your cookies file, as the application will automatically search for it. And if you want to use an ignorelist to filter words but the file doesn't exist, Synthalingua will offer to create a template for you.

This has been a massive undertaking, and I'm focused on making sure everything is as stable as possible before release. I'm aiming to get this update out to all of you soon!

Thank you for your incredible support!

- Cyberofficial