Whisper.cpp is underwhelming

<!-- SC_OFF --><div class="md"><p>I've been running openai-whisper installed via pipx for some while now, using it to transcribe voice notes taken via a digital recorder, and

<!-- SC_OFF --><div class="md"><p>Did not impress me much. Even using tags, 90% audio comes out as robotic TTS. Weird emotionless audio.<br /> And it's not really open source as

<h2> Whisper Hallucination on Silence: Why Your Transcript Loops the Same Phrase </h2> <h3> The Problem </h3> <p>Have you ever transcribed audio with Whisper (or any ASR model)

<!-- SC_OFF --><div class="md"><p>I am running Faster Whisper on CPU only and get good running times with about 2.5 min for 60 min sound with Whisper Base. With Pyannote for

<!-- SC_OFF --><div class="md"><p>I want to do voice transcription with AI using models like Nvidia Whisper Large Model, which has MLX variants for apple silicon.</p> <p>Whats the

<!-- SC_OFF --><div class="md"><p>I am working with transcriptions and faster-whisper (also have tried whisper.cpp earlier) and the context provided does not really have any