TranscriptionSuite Bundles Whisper, NeMo and VibeVoice-ASR Into a Local, GPU-Accelerated Speech-to-Text App With Speaker Diarization

A new open-source project called TranscriptionSuite packages several leading speech-recognition engines into a single, fully local application that transcribes audio and identifies who is speaking, all with GPU acceleration and no cloud dependency. The tool, hosted on GitHub under the homelab-00 account, aims to make private, on-device transcription with speaker diarization accessible across Windows, macOS and Linux.