Hosts and guests separated
With a plan, MAI-Transcribe-2 detects who is speaking — even with three or four voices.
Show notes · Blog · Accessibility
Transcribe your podcast
A transcript makes your episode findable and accessible: search engines cannot read audio, and neither can deaf listeners. Upload the episode and get text with speakers and timestamps — the basis for show notes, chapters and blog posts.
Free trial: 3 transcriptions per calendar month, each up to 10 minutes, no account — as plain text. Speaker separation (“who said what”) and files up to 4 hours are part of the paid offerings: day pass €2.99, credit €4.99, Pro €8.99/month.
Upload an episodeHow it works
- Step 1
Export the episode
Mono at 64 kbps is enough for recognition: about 29 MB per hour, so long episodes stay under the 100 MB limit.
- Step 2
Upload or link
Upload the file or paste the direct MP3 link from your RSS feed.
- Step 3
Reuse it
TXT for show notes, Word for the blog post, SRT or VTT for the YouTube video.
What you get
Timestamps for chapters
Every passage has a timestamp, so topic changes are easy to find and turn into chapters.
Up to 4 hours
Long conversation episodes run in one go.
Subtitles included
For video podcasts, you get SRT and VTT, with speaker names if you like.
At a glance
- Direct links from your RSS feed work as long as the file is public and under 100 MB.
- Music intros and jingles are not recognized as speech and usually do not appear in the text.
- Pro (€8.99/month) pays off for a weekly show; for single episodes, credit is enough.
- Free trial without an account or login; plans from €2.99 one-off, no subscription required.
- Technical anti-abuse protection: currently up to five transcription starts per hour and files up to 100 MiB.
- Processed via OpenRouter with zero data retention: the AI providers store neither audio nor text; Sprechverlauf deletes audio files after processing and keeps transcripts only for signed-in users — encrypted.
- Premium: upload videos (MP4, MOV, WebM, up to 100 MiB), name speakers and review subtitles before exporting SRT or VTT.
- The video is deleted from our server one hour after upload. Your encrypted transcript stays in your account until you delete it. Only the extracted audio is sent for transcription.
Frequently asked questions
Does a transcript help my podcast on Google?+
Yes, indirectly: Google indexes text, not audio. A transcript or an article based on it on the episode page supplies the words people search for.
My episode is larger than 100 MB. What now?+
Export a copy in mono at 64 kbps. That is plenty for speech recognition, and one hour takes only about 29 MB.
How well does it work with several guests?+
Well, as long as people don't all talk at once. Separate tracks or at least separate microphones improve speaker separation a lot.
Can I publish the transcript as is?+
Read it once first: names, brands and jargon are the usual errors. Click a timestamp to hear the passage.
Do I need an account or a subscription?+
Not for the free trial. You only need an account to buy a plan (day pass, credit or Pro) — a subscription is never required. Technical anti-abuse limits apply.
How long is processing data retained?+
Recognition runs through OpenRouter with zero data retention — the providers (Microsoft Azure, DeepInfra, Groq) store neither audio nor text. Sprechverlauf deletes uploaded audio files after processing, at the latest after two hours, and keeps the transcript in memory for at most two hours. Only signed-in users get an encrypted copy in their account, which they can delete at any time.
Video uploads, speaker names and subtitles+
Premium: upload videos (MP4, MOV, WebM, up to 100 MiB), name speakers and review subtitles before exporting SRT or VTT. The video is deleted from our server one hour after upload. Your encrypted transcript stays in your account until you delete it. Only the extracted audio is sent for transcription.