SpeakerScope v1.0.0 - Hackathon Release

SpeakerScope v1.0.0

Live speaker-attributed transcription for Windows - native C++ desktop app built for the AssemblyAI x lablab.ai hackathon.

• AssemblyAI Universal Streaming (v3 WebSocket) for real-time word-level transcription
• NVIDIA Nemotron-3-Diarization on-device via audio.cpp (GGUF q8_0, CPU-only - model bundled)
• Live open-turn speaker labels, overlap detection, editable Person 1-8 lanes, per-person filter/export
• 117/117 automated tests; headless sim + benchmark tools included

Run it

1. Unzip speakerscope-win-x64.zip
2. Copy .env.example to .env and set ASSEMBLYAI_API_KEY=<your key>
3. Run speakerscope.exe - pick mic / system loopback / WAV and press Start