Formerly NovaScribe — same team, same product, refreshed name. Read the announcement →
Podcast Transcription
Upload an episode, paste a YouTube link, or drop in the audio URL. You get a transcript with the speakers separated, exports to TXT, DOCX, SRT and VTT, and a file you actually own.
Drop your file here
or click to browse
Audio: MP3 · WAV · M4A · FLAC · OGG · AAC · AIFF · WMA · AMR · OPUS
Video: MP4 · MOV · AVI · MKV · WebM · FLV · WMV
Up to 200 MB · 5 min free preview
Free preview covers the first five minutes — enough to hear whether it separates your speakers correctly, not a whole episode. Results have a Transcript / SRT toggle.
A 60-minute episode is roughly 9,000 words of speech that exists nowhere as text until someone transcribes it. VexaScribe takes the audio file, a YouTube link, or the episode's direct audio URL and returns a timestamped transcript with speakers separated, plus exports to TXT, DOCX, SRT and VTT. Median processing time is about 3 minutes per hour of audio on Standard and about 1 on Premium, measured across 5,395 production jobs in October 2026. Check what you already have first: Apple Podcasts auto-generates transcripts in eleven languages, and Spotify does it by default for shows hosted on Spotify for Creators. If a searchable transcript inside those apps is all you need, you are done. What the platforms do not give you is a file you own, real speaker names instead of generic labels, SRT for a video cut, bulk processing of a back catalogue, or text you can publish on your own site where search engines and AI assistants can cite it. Free tier covers 30 minutes; paid plans start at $2/month.
How to get a transcript of your own podcast episode
Three ways in, depending on where the episode already lives. All three produce the same thing: a timestamped transcript with speakers separated, exportable as TXT, DOCX, SRT or VTT.
1. Upload the audio file
The master file from your editor, or whatever your host lets you download — MP3, WAV, M4A, MP4, MOV. This is the best-quality route because you are handing over the original audio rather than a re-encoded stream.
2. Paste the YouTube URL
If you publish a video version, paste that link and skip the download entirely. Useful for back-catalogue work where the video is the only copy you can still get at.
3. Paste the episode's direct audio URL
Every podcast RSS feed exposes a direct MP3 link for each episode — the <enclosure> URL. Open your feed, copy the link for the episode, paste it in. This works for any show with a public feed, including your own, without downloading anything first.
What does not work, and why: Spotify links. Spotify's own transcripts are deliberately not exportable, and we do not scrape them. If Spotify is the only place your episode exists, use the audio file from your host instead — every podcast host gives you one.
How VexaScribe Transcribes Your Podcast
Three steps from upload to finished transcript. No software to install.
- 1
Upload your episode
MP3 (typically 28-56 MB/hr at 64-128 kbps), WAV, M4A, MP4 video. Up to 5 GB. Most podcast episodes are well under this.
- 2
AI transcribes with speaker labels
Whisper Large-v3 transcribes; diarization automatically separates speakers (Speaker 1, Speaker 2, Speaker 3). Rename to "Host", "Guest Name", actual names — labels carry through every export.
- 3
Export everything you need
Plain text for show notes, DOCX for editing, SRT/VTT for video versions, JSON for developers, AI summary with chapters and key topics.
What You Get From One Upload
Everything you need to ship an episode — show notes, subtitles, blog content, multilingual versions.
Speaker-labeled transcript
Auto-diarization separates host, co-host, and guests. Useful for interview shows and panel formats. Rename speakers to actual names in the editor.
AI-generated show notes draft
Chapter markers, key topics, episode summary. Use as a starting point for your show notes — saves 30-60 min of manual writing per episode.
SRT/VTT subtitles
For YouTube, Vimeo, or video podcast versions. Word-level timestamps; perfectly synced. Upload directly to YouTube using "Add language → Upload file with timing".
Translated transcripts
Optional: translate to any of 133 languages using built-in translation. Reach international audiences without re-recording — Spanish podcast → English subtitles in one workflow.
Why Podcasters Transcribe
Four specific business reasons. Most podcasters end up with all four, not just one.
1. SEO
Search engines cannot index audio. A 60-minute episode adds ~9,000 words of indexable text, and it compounds across a back catalogue. The mechanism is specific: a transcript published on your own site creates a page that can rank for the long-tail phrases your guests actually said — product names, techniques, arguments — language you would never think to put in an episode description. It only works if the text lives somewhere you control; a transcript inside Apple or Spotify is not indexable by anyone else. This American Life reportedly saw a 4.36% inbound traffic lift after transcribing their archive (2014, single datapoint — directional only, and we have not replicated it).
2. AI search citation
ChatGPT, Perplexity, Claude, and Google AI Overviews can only cite text. If your episode discusses a topic but exists only as audio, AI engines won't surface it. Transcripts are the only path for podcast content to appear in AI answers.
3. Accessibility
WCAG 2.1 SC 1.2.1 (Level A) requires text alternatives for prerecorded audio. Public-entity podcasts will fall under ADA Title II — deadline extended to April 26, 2027 for entities serving 50,000+ people, April 26, 2028 for smaller. Private podcasts aren't legally required, but transcripts are still expected by deaf and hard-of-hearing listeners.
4. Content repurposing
One episode → show notes (typically 1,500–2,500 words from a 45–60 min episode) → blog post → social clips → newsletter → quote graphics. Most repurposing tools (Castmagic, Riverside Magic Clips, Opus Clips) start from a transcript.
Speaker Identification: Hosts, Co-hosts, Guests
VexaScribe automatically detects different speakers in your podcast and labels them 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.
How well speaker separation works, by recording setup
Separating speakers is measured in Diarization Error Rate — the share of audio attributed to the wrong person, so lower is better. We have not run our own DER benchmark, so the ranges below are the published figures for the task itself, the same ones behind our speaker diarization comparison. They describe how hard the problem is under each setup rather than how any one vendor scores. Most interview podcasts sit in the top two rows.
| Recording setup | DER | Notes |
|---|---|---|
| Solo host, one mic | n/a | Nothing to separate |
| 2–3 people, separate mics, remote-recorded | 5–8% | Best case — most interview podcasts |
| 2–4 people, standard conditions | 7–12% | What consumer benchmarks measure |
| 5+ speakers | 12–18% | Degrades as speakers are added |
| Shared mic in one room | 15–25% | Panel shows recorded around a table |
| Heavy crosstalk and interruption | 30%+ | DIHARD III shows this is the failure mode |
The practical read: if your guests record on their own microphones — the normal remote-interview setup — speaker separation is close to a solved problem. If everyone shares one mic around a table, or people talk over each other constantly, expect to fix some labels by hand. Renaming a speaker once updates every instance of it, so even the bad case is minutes of work rather than hours.
State-of-the-art research benchmarks show pyannote 3.1 at 11–19% Diarization Error Rate (DER) on real-world audio. PyannoteAI commercial: 11.2% DER. (Sources: arXiv 2509.26177, picovoice 2025 benchmark.)
Recording on Riverside.fm or similar? Their separate-track recording (each speaker as their own uncompressed 24-bit/48 kHz WAV file) gives you near-perfect speaker ID — diarization isn't even needed because you've already separated speakers at the recording stage. Source: riverside.com.
Cost: AI vs Human vs Self-Transcribing
For most podcasters the question is not a percentage, it is whether a listener searching the transcript lands on the right moment. Podcasts are discovery and accessibility content, not legal records — a guest's name spelled wrong is a five-second fix, not a failed deliverable.
| Method | Cost per hour | Time to deliver |
|---|---|---|
| VexaScribe (AI) | $0.20–$0.60/hr depending on plan | 3–5 min per audio hour |
| Otter.ai (AI) | From $16.99/mo | Real-time |
| Sonix (AI) | ~$10/hr pay-as-you-go | 5–15 min |
| Rev AI (AI) | $15/hr ($0.25/min) | 5–10 min |
| Rev (human) | ~$119/hr ($1.99/min) | 12 hours or less |
| Happy Scribe (human) | ~$102/hr ($1.70/min) | 4–24 hours |
| Self-transcribe | Your time (~4 hrs typing per 1 hr audio) | Days |
The cost difference between AI ($0.20–$0.60/hr on VexaScribe) and human ($90–$150/hr) is 150–450×. Most podcasters use AI and edit any errors manually (proper nouns, brand names, technical jargon).
From Transcript to Show Notes in 5 Minutes
A typical 45–60 minute interview produces around 7,000–9,000 words of transcript. Manually writing show notes from that takes 30–60 minutes. VexaScribe's AI summary feature does it automatically — you get:
- ✓Episode summary (2-3 paragraphs)
- ✓Chapter markers with timestamps (great for YouTube and Apple Podcasts chapters)
- ✓Key topics discussed
- ✓Notable quotes with timestamps
The AI output is a starting point, not final copy. Most podcasters edit for brand voice, add their own takes, and pull specific quotes manually. But it cuts the writing time roughly in half.
Beyond summaries: AI Chat lets you ask the transcript questions directly — “pull three punchy quotes I can use for social clips,” “what did the guest say about pricing?”, “find the strongest soundbite about productivity.” Answers come back with clickable timestamps that jump the audio player to the exact moment. Available on paid plans.
Chapter Markers for Spotify, Apple Podcasts & YouTube
Chapters help listeners jump to segments they care about (interview start, main topic, Q&A, sponsor break) and reduce drop-off. Most podcasters write chapters by hand from listening back — 10–20 minutes per episode. VexaScribe generates chapter markers with timestamps from the transcript in a single click.
The workflow
- Upload episode → get transcript (5–10 min)
- Generate AI summary with the Podcast template — chapter markers included in the output
- Copy timestamps and titles into your podcast host's chapter editor
Where chapters go, by platform
| Platform | How chapters get in |
|---|---|
| Apple Podcasts | ID3 chapter tags embedded in episode MP3 (via your host's chapter editor) or Apple Podcasts Connect chapter fields |
| Spotify | Timestamps in episode description auto-parse into clickable jump points; for full chapters, use RSS with <podcast:chapters> tag (Podcasting 2.0) |
| YouTube (video version) | Timestamps in video description auto-create chapter cards on the progress bar; must start with 00:00 and have ≥3 chapters |
| Buzzsprout / Transistor / Simplecast / Captivate | Paste chapters into the episode edit page; the host writes ID3 tags on your behalf |
VexaScribe outputs chapter timestamps in a copy-paste-friendly plain-text format that works with all the destinations above. The Podcasting 2.0 <podcast:chapters> JSON format is on the roadmap.
Bulk Transcription for Backlogs and Agencies
Prolific podcasters with an un-transcribed back catalog, and agencies servicing multiple shows, need to process dozens of episodes at once. VexaScribe supports up to 50 episodes per batch — upload once, walk away, come back to a ZIP archive with every transcript plus AI summaries.
- ✓50 files per batch upload, mixed audio and video formats
- ✓ZIP export of all transcripts + AI summaries in one download
- ✓Same plan minutes — no bulk premium; the Studio plan at $20/mo covers ~100 hours (roughly 100 one-hour episodes)
- ✓Independent per-file processing — one failing file doesn't block the batch
For very large backlogs (100+ episodes), split into multiple 50-file batches back-to-back. See the bulk transcription feature page for the detailed workflow.
What 100 hour-long episodes actually costs
Worth doing the arithmetic before you start, because a back catalogue is the one case where the numbers get large enough to matter:
- •Minutes needed: 100 × 60 = 6,000 minutes.
- •Cost: exactly one month of Studio at $20 — 6,000 minutes is the Studio allowance. That works out to $0.20 an episode.
- •Processing time: at the measured median of 3–5 minutes per audio hour, 100 hours is roughly 5–8 hours of machine time — two batches of 50, started and left alone. Files process independently, so transcripts arrive as they finish.
- •The same job through Rev at $1.99/min would be about $11,940 (verified 19 September 2026).
The honest caveat: you are not going to review 100 transcripts. Budget review time for the episodes that matter — the ones you will quote, republish or cut clips from — and let the rest sit as searchable archive text, which is most of the value anyway.
If You Have a Video Version: SRT Subtitles
YouTube version of your podcast? Spotify Video? VexaScribe exports SRT and VTT subtitle files with word-level timestamps from the same transcription. Upload directly to YouTube using Add language → Upload file with timing → select your .srt file.
Platforms that accept SRT/VTT for podcasts
- ●Buzzsprout — TXT, SRT, VTT
- ●Transistor — SRT, VTT, TXT, JSON
- ●Spotify for Creators — VTT, SRT
- ●Apple Podcasts — SRT/VTT via RSS <podcast:transcript> tag
- ●YouTube — SRT
For the full SRT workflow, see the SRT generator page.
Transcribing Non-English and Multilingual Podcasts
VexaScribe transcribes in 99 languages. Important caveat about Apple Podcasts: their auto-transcripts only support eleven languages (English, Danish, Dutch, Finnish, French, German, Italian, Norwegian, Portuguese, Spanish and Swedish) (as of iOS 17.4 launch in March 2024). Spanish, French, German podcasters can rely on Apple's auto-transcripts; everyone else needs to provide their own.
Transcripts can also be translated to any of 133 languages using VexaScribe's built-in translation. Useful for international episodes — record in Spanish, publish English-translated transcript and SRT for English-speaking audiences.
See the transcribe and translate workflow →What it costs for a real publishing schedule
Work out your minutes before you pick a plan. A weekly hour-long show is 4.33 episodes a month, about 260 minutes — not four hours, which is the rounding that trips people up.
| Plan | Price | Included | Per hour | Fits |
|---|---|---|---|---|
| Free trial | $0 | 30 min total | — | One episode, no card |
| Starter | $2/mo | 200 min | $0.60/hr | Biweekly hour-long, or weekly under 45 min |
| Basic | $5/mo | 1,000 min | $0.30/hr | Weekly hour-long show, with room to spare |
| Pro | $10/mo | 2,500 min | $0.24/hr | Several shows, or a backlog in chunks |
| Studio | $20/mo | 6,000 min | $0.20/hr | 100 hours a month — agencies and networks |
The short version. A weekly hour-long show needs Basic at $5/month, which leaves 740 minutes spare for bonus episodes or a backlog. A biweekly show, or a weekly show under 45 minutes, fits Starter at $2. An earlier version of this page recommended Starter for a weekly hour-long show; that was wrong — 260 minutes does not fit in 200 — and it is corrected here. For comparison, the same 4.33 hours through Rev's human transcription at $1.99/min would be about $517 (verified 19 September 2026).
Frequently Asked Questions
How do I transcribe a podcast episode?
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 3-5 minutes.
Does podcast transcription identify different speakers?
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.
What can I do with my podcast transcript?
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.
How long does it take to transcribe a 1-hour podcast?
A typical 1-hour podcast episode is transcribed in about 3-5 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.
What audio formats work for podcast transcription?
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.
Can I transcribe podcasts in different languages?
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.
How much does podcast transcription cost?
Work it out in minutes. A weekly hour-long show is 4.33 episodes a month, about 260 minutes — that is Basic at $5/month (1,000 minutes), not Starter at $2 (200 minutes). A biweekly show, or a weekly show under 45 minutes, fits Starter. Plans run $2 for 200 minutes, $5 for 1,000, $10 for 2,500 and $20 for 6,000, which works out between $0.60 and $0.20 per audio hour. The same 4.33 hours through Rev’s human transcription at $1.99/min would be roughly $517 (verified 19 September 2026).
Can I export subtitles for YouTube from my podcast transcript?
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).
Does Apple Podcasts already transcribe my show automatically?
Apple Podcasts launched auto-generated transcripts in March 2024 with iOS 17.4 — but only for eleven languages (English, Danish, Dutch, Finnish, French, German, Italian, Norwegian, Portuguese, Spanish and Swedish). 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.
How accurate is AI transcription for podcasts with multiple speakers?
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.
Can I generate chapter markers for Spotify and Apple Podcasts from my transcript?
Yes. Chapter markers with timestamps are one of the outputs when you generate an AI summary from your transcript — pick the Podcast summary type in the summary editor to get a chapter list with times, per-chapter descriptions, and key quotes. Apple Podcasts and Spotify both accept chapter markers via ID3 chapter tags embedded in the episode audio file. Most podcast hosting platforms (Buzzsprout, Transistor, Captivate, Simplecast) let you paste chapter timestamps directly into the episode edit page rather than encoding them into the audio. Chapters help discovery on both platforms and reduce listener drop-off.
Can I transcribe a backlog of old podcast episodes in bulk?
Yes. VexaScribe supports bulk transcription — upload up to 50 episodes at once as a single batch, and download all transcripts plus AI summaries as a ZIP archive when the batch completes. This is the workflow most podcast agencies and prolific podcasters use to catch up on 6-12 months of un-transcribed back catalog in one session. Batch processing uses the same $2-$20/month plan minutes as single uploads. For very large backlogs (100+ episodes), split into multiple batches; the Studio plan at $20/month covers about 100 hours of audio.
Can I get a full SEO blog post from a podcast episode transcript?
Yes, in two steps. First, get the transcript. Second, use the Podcast summary type to generate structured show notes with chapters, key quotes, and speaker discussion points — that's already usable as a blog post skeleton. For a longer piece, use the transcript editor's regenerate feature with the Lecture or General summary type to get a different structural angle, then paste both into your editor of choice. A typical 45-60 minute episode produces 1,500-2,500 words of blog content. Publishing the transcript itself alongside the episode also has meaningful SEO impact — searchable long-tail keywords the audio can't rank for on its own.
How do I get a transcript of any podcast?
Five methods, honest order. (1) Apple Podcasts — since March 2024, most English/French/Spanish/German shows have auto-transcripts. Open the episode, tap the quote-bubble icon at the top-right. Free, no signup, viewer-only (no download). (2) Spotify — creators can opt-in to display transcripts on the episode page; coverage is thin outside major shows. (3) HappyScribe podcast library (podcasts.happyscribe.com) — user-uploaded searchable transcripts, mostly popular English shows. (4) YouTube — if the podcast publishes to YouTube, use /tools/youtube-transcript to pull the auto-caption. (5) Upload the audio yourself to VexaScribe — works for any podcast, any language, produces a downloadable transcript in TXT/DOCX/SRT with speaker labels. Method 5 is the only path that reliably works for non-English shows, obscure podcasts, and any use case that needs the transcript as a file.
Where can I find existing podcast transcripts online?
Four places. (1) Apple Podcasts app — tap the quote-bubble icon on episodes that have auto-transcripts (mostly English/French/Spanish/German post-2024). (2) Spotify — some shows opt-in to display transcripts on the web episode page; check the episode's page on open.spotify.com. (3) HappyScribe's podcast library (podcasts.happyscribe.com) — searchable database of user-contributed podcast transcripts, tilted toward popular English shows. (4) The podcast's own website or Substack — some hosts publish transcripts for SEO or accessibility. If none of these have your show, upload the episode audio here and get a full transcript back in 3-5 minutes.
Does Spotify have podcast transcripts?
Partially, since 2023. Spotify supports creator-uploaded transcripts (VTT/SRT) that display on the web episode page and inside the app on some devices. It's opt-in, not automatic — the creator has to upload the file — so coverage is thin outside major podcasts. Spotify's own official podcasts have transcripts; smaller independent shows usually don't. If Spotify doesn't show a transcript for the episode you want, download the audio (Spotify allows for offline listening on paid plans; RSS-published shows are downloadable directly) and upload here for a full AI transcript in your language.
Does Apple Podcasts have transcripts?
Yes, for many shows since March 2024 (iOS 17.4). Apple auto-generates transcripts for episodes in eleven languages (English, Danish, Dutch, Finnish, French, German, Italian, Norwegian, Portuguese, Spanish and Swedish) — no creator action required. Look for the quote-bubble icon (💬) in the top-right of the episode player; tap it to view the transcript. Limits: (1) viewer-only, no download or export; (2) English/French/Spanish/German only — other languages don't have auto-transcripts yet; (3) accuracy is Apple's proprietary model, comparable to Otter, not always as accurate as Whisper Large-v3. For non-Apple-supported languages, exportable files, or higher accuracy, upload the audio here.
Related
YouTube transcript downloader
If your podcast is on YouTube too — paste the URL, download captions
Transcribe audio to text
All formats, 99 languages, 95% accuracy
MP3 to text
Most podcasts are MP3 — convert to text in 2 minutes
WAV to text
Studio masters from your DAW — straight to transcript
Best podcast transcription tools
Honest comparison of 10 tools — Descript, Castmagic, Otter, Rev, more
Bulk transcription
50 episodes per batch — for podcast agencies and back-catalog projects
Sermon transcription
Same single-speaker pattern — workflows for pastors and churches
Medical transcription
For medical podcasts, CME content, conferences — when general AI is appropriate
Transcript formatting
SEO-friendly podcast transcript format — H2 sections, speaker labels
Best subtitle generators 2026
12 tools compared — for audiogram subtitle production
Speaker labels — how they work
Multi-host podcast labeling — pipeline, accuracy, cross-file rename
Transcript to summary
Generate show notes & chapters from any transcript
SRT generator
Subtitle files for video podcasts
Video to SRT
Workflow for video podcast SRT generation
Video to text
Plain-text transcript from any video podcast file
M4A to text
For guest interviews recorded on iPhone Voice Memos
OGG to text
For guest interviews via WhatsApp voice notes (Android)
Transcribe Spanish audio
Spanish-language podcast workflows + EN translation
Captions vs subtitles
Picking the right caption type for video podcasts
How to add subtitles to a video
Step-by-step for YouTube, Premiere, CapCut, iPhone
Interview transcription
Workflow for guest-interview podcasts + speaker labels
Transcribe and translate
Reach international audiences in 133 languages
How accurate is Whisper?
WER benchmarks by language and condition
Whisper transcription
Hosted Whisper Large-v3 — the engine behind podcast transcripts