Releases: santhoshkumar-dev/ghostly
1.0.0
🎉 Ghostly v1.0.0
Ghostly now features a powerful, completely local Live Audio Transcription Engine, allowing it to listen to your interviews in real-time and provide AI-generated context without taking screenshots!
✨ New Features
🎧 Live Interview Copilot
Added a sleek, minimal inline audio panel to the main UI.
It doesn't block your screen or mouse, allowing you to seamlessly use it alongside your coding environment.
🔊 Dual-Channel Audio Capture
Simultaneously records:
- Microphone (you)
- System Loopback Audio (interviewer)
Both streams are captured in real-time.
🤖 On-Device AI Transcription
Uses @xenova/transformers to run the Whisper Tiny model locally inside a Web Worker.
Benefits:
- 🔒 Absolute privacy
- ⚡ Zero latency
- 💻 Fully on-device
🎙️ Voice Activity Detection (VAD)
Intelligently chunks audio streams.
When you or the interviewer pause briefly, the system instantly:
- Detects the pause
- Transcribes the previous sentence
🚀 Copilot Integration
Send the live ongoing transcript directly to your active LLM provider with the new:
Ask Copilot 🚀 button
🛠️ Fixes & Improvements
🐞 Desktop Audio Crash Fix
Fixed an issue where the renderer would crash when capturing desktop audio on:
- Windows
- macOS
🧠 Transformers.js Transcription Fix
Resolved an issue where Transformers.js failed to transcribe audio due to Float32Array serialization loss across the Web Worker boundary.
🔎 Built-in Audio Debugger
Developer logs now include raw WAV playback of what the AI actually hears, making it easier to debug microphone issues.
🍎 macOS Build Fix
Resolved a .dmg naming conflict during the GitHub Actions build pipeline that caused macOS artifacts to fail.
📦 Download
Mac (Apple Silicon / M1, M2, M3)
📥 Download ghostly-1.0.0-arm64.dmg
Mac (Intel)
📥 Download ghostly-1.0.0-x64.dmg
Windows
📥 Download ghostly-1.0.0-setup.exe