Podcast Transcription Service

Depends on your workflow. For most independent podcasters and small networks who want a clean transcript with multi-host speaker labels, VexaScribe gives 30 minutes free on signup, then $2–$20/month for higher volume — at the $20 Studio tier, that works out to roughly $0.20 per hour of audio. Otter has a generous free tier (300 min/month) but is meeting-recording-first; Descript is excellent if you also want a podcast editor in the same tool (different product category — they own that space); Riverside bundles recording + transcription at $24+/month if you also need to record remotely. Rev's human transcription is the most accurate (~99%) but costs ~$90 per 1-hour episode — only worth it for high-stakes work. For pure cost-per-minute at scale, install OpenAI Whisper locally and pay $0.

Yes — and most podcasters don't realize the size of the opportunity. Podcast audio is invisible to Google search by default; the only thing search engines can index is your episode title and description. Publishing the transcript turns every spoken word into searchable text. Pacific Content and Edison Research have repeatedly documented that shows publishing full transcripts see meaningful organic traffic lift compared to audio-only shows — typical reports range from 2–5× organic search growth over 6–12 months. Bonus: accessibility (the CDC estimates ~15% of US adults have some degree of hearing loss) and international audience (transcripts can be translated, audio can't).

Whisper-based speaker diarization (which VexaScribe uses) is most accurate with 2–4 distinct voices. Realistic accuracy by speaker count: 2 speakers (typical solo + guest) → 95%+ label accuracy; 3–4 speakers (host + 2–3 guests) → 90–95%; 5–6 speakers (panel format) → 80–90%; 7+ speakers (chaotic roundtables) → requires manual cleanup. The hardest cases: same-gender voices with similar tone, and any segment with overlapping speech. Best practice for podcasters: after the first transcription pass, rename "Speaker 1" → host name, "Speaker 2" → guest name, then save the named pattern for future episodes.

Yes — long-form is increasingly common (Joe Rogan, Tim Ferriss, Lex Fridman, Acquired all run 2–4+ hour episodes). VexaScribe processes long episodes as a single file with no need to split. Realistic timing: 1-hour episode ≈ 5–10 min to process; 2-hour ≈ 15–20 min; 3-hour ≈ 25–30 min. File size cap is 5 GB per upload, which covers roughly 6 hours of high-quality 256 kbps MP3 or about 4 hours of 1080p video podcast. Most free transcription tools cap at 25 MB (~30 minutes of audio) — a real constraint for the long-form format.

Yes. Upload the MP4/MOV directly — VexaScribe extracts audio internally. No need to convert. If your video podcast lives on YouTube and you don't have the source file, our YouTube transcription tool accepts video URLs directly. For Riverside recordings (high-quality WAV + MP4), use either file. The transcript output is the same; the SRT export is useful if you're also uploading the video to YouTube/Vimeo and want captions.

About 10–15% of audio length on AI tools: 1-hour episode → ~5–10 min, 2-hour → ~15–20 min, 3-hour → ~25–30 min. Processing happens server-side, so you can close the browser tab and come back when it's done — the transcript saves automatically. Human transcription services (Rev, GoTranscript) take 4–24 hours regardless of episode length.

They're different deliverables for different jobs. A transcript is the full literal text of everything said in the episode — typically 8,000–15,000 words for a 1-hour interview, mostly used for SEO (publishing on your website), accessibility, content repurposing, and research. Show notes are a curated summary: 2–4 paragraphs of highlights, a timestamped list of topics or chapter markers, and links to anything mentioned — typically 300–800 words, written FROM the transcript after the episode is done. VexaScribe produces the transcript; show notes are something you write (or generate with our summary tool) from the transcript as raw material.

Yes — for personal research, journalism, academic study, competitive analysis, or quote sourcing, transcribing a podcast you don't own is generally fine under fair-use principles (specifics vary by jurisdiction). You can either upload an MP3/MP4 of the episode you've saved locally, or use our YouTube transcription tool if it's a video podcast on YouTube. For commercial republishing (e.g., publishing transcripts of someone else's podcast on your own site as content), you'd need permission from the podcast creator — the transcript itself can be a derivative work for copyright purposes.

Audio formats from any podcast host or recording app: MP3 (Buzzsprout, Anchor, Libsyn exports), WAV/AIFF (studio sessions in Hindenburg, Pro Tools, Audacity, Reaper), M4A (iPhone/QuickTime field recordings), FLAC, OGG, AAC. Video formats for video podcasts: MP4, MOV, MKV, WEBM. Mix audio + video in the same workflow without conversion — VexaScribe handles audio extraction automatically.

For 100 typical 1-hour podcast episodes (~6,000 minutes total), the math: VexaScribe Studio at $20/month covers it — that's roughly $0.20/hour or $0.003/minute, all-in flat pricing. Deepgram API is roughly $0.22/hour pay-as-you-go at base rates ($22 for the full batch) but requires developer setup. Rev human transcription at $1.25–$1.99/min would cost $7,500–$11,940 for the same 100 episodes — only worth it if you need legal-grade accuracy. Whisper installed locally is $0 if you have a GPU machine and the patience for batch scripting. For most podcasters with a back-catalog, VexaScribe Studio for one month is the simplest path. See our bulk transcription page for the parallel-upload workflow.