mercuryScribe vs Transcript.so

mercuryScribe is a locally run web interface for transcribing audio and video, with transcript editing and document export. It uses Faster Whisper for speech transcription and pyannote models to distinguish speakers. Users can edit transcripts in a browser, save changes as VTT files, and generate HTML or DOCX files containing speaker names and timestamps. The FAQ says audio, transcripts, and metadata are not sent to external services; the models are obtained from Hugging Face. Users need a Hugging Face access token and must accept terms for the required models. MAXQDA support is included by default, and custom integrations can be developed. The maker documents Docker, Windows package, and Python package installation options. The software is free and open source under the MIT license for personal, academic, or commercial use, and is listed for Linux, macOS, Windows, and web. The maker describes it as built for researchers and also suitable for journalists, podcasters, organizations, educators, accessibility groups, and professionals. Professional services include installation help, training, workshops, and private server setup with GPUs.

Transcript.so converts uploaded audio and video into searchable text. It detects languages automatically, labels speakers, lets users click a sentence to replay that moment, and exports transcripts as TXT, DOCX, PDF, CSV, SRT, or VTT. The service supports more than 90 languages and accepts formats including MP3, WAV, M4A, AAC, FLAC, OGG, OPUS, WMA, AIFF, MP4, MOV, WEBM, MKV, MPEG, and AVI. Users can search saved transcripts for topics. The maker says Whisper handles most files and another diarization engine is selected for multi-speaker or long recordings. The free plan allows three transcriptions daily, with files up to 30 minutes or 50MB. Unlimited costs $10.00 per month billed yearly ($120). Accuracy may drop with names, jargon, crosstalk, strong accents, or noise. Files remain until deleted and are encrypted, private, and not used to train models.