Podcast Transcription

Upload your podcast episode file (MP3, WAV, M4A, or other supported formats) to VexaScribe (formerly NovaScribe). The AI automatically transcribes the audio using OpenAI's Whisper Large-v3, detects different speakers (host, co-host, guests), and generates word-level timestamps. A typical 60-minute episode completes in 5-10 minutes.

Yes. VexaScribe includes automatic speaker diarization. When your podcast has multiple speakers, the system labels each one separately (Speaker 1, Speaker 2, Speaker 3...). You can rename them in the editor ("Host", "Guest: Jane Smith", "Co-host: Dan") and the labels carry through every export. Accuracy is highest for 2-3 speakers on separate microphones; degrades with overlapping speech and 4+ speakers.

Turn one episode into multiple pieces of content: create show notes for your website, repurpose as blog posts (1,500-2,500 words from a 45-60 min episode), extract quotes for social media, generate YouTube captions (SRT/VTT export), improve SEO with searchable text, and make your content accessible to deaf and hard-of-hearing listeners.

A typical 1-hour podcast episode is transcribed in about 5-10 minutes. Processing time depends on audio length, quality, and current load. You can close your browser while processing — the transcript will be waiting in your dashboard when it's ready.

VexaScribe supports all common podcast formats: MP3, WAV, M4A, FLAC, OGG, AAC, AIFF, WMA, AMR, OPUS for audio. Video formats (MP4, MOV, AVI, MKV, WebM) work too — we extract the audio automatically. Files can be up to 5 GB and 10 hours per upload, which covers virtually any podcast episode.

Yes. VexaScribe supports 99 languages for transcription with automatic language detection. You can also translate transcripts to any of 133 languages using the built-in translation feature — useful for international podcasts that want English subtitles, or English podcasts reaching non-English audiences.

VexaScribe plans start at $2/month for 200 minutes. A weekly 1-hour podcast (~4 hours/month) costs less than $3/month on the Starter plan. Compared to human transcription services like Rev ($1.50/min, ~$90/hr) or Happy Scribe ($1.70/min), AI is 150-450× cheaper. For most podcasters, AI accuracy at ~95% on clear audio is more than enough.

Yes. VexaScribe exports SRT and VTT subtitle files with word-level timestamps. Upload directly to YouTube using "Add language → Upload file with timing" → select your .srt file. Same files work in Vimeo, Spotify Video, or any video editor (Premiere Pro, Final Cut Pro, DaVinci Resolve).

Apple Podcasts launched auto-generated transcripts in March 2024 with iOS 17.4 — but only for English, French, Spanish, and German. Auto-transcription rolled out to new episodes first, with the back catalog filling in over time. If you publish in another language, or want to control the exact transcript shown, you'll still need to provide your own. Spotify also added optional creator-uploaded transcripts (VTT/SRT) starting in 2023-24, but it's opt-in, not automatic for all podcasts.

On clear podcast audio with 2-3 speakers on separate microphones, AI transcription accuracy is around 95% (5% Word Error Rate). Speaker diarization is accurate when speakers have distinct voices and don't talk over each other. State-of-the-art research shows pyannote 3.1 achieves 11-19% Diarization Error Rate on real-world audio. Conditions that hurt accuracy: overlapping speech, similar voices, 4+ speakers, low-bitrate audio.