Konthora

Knowledge Center

How to Transcribe a Podcast

A podcast transcript does three jobs at once: it makes your episode searchable, it lets listeners skim instead of listen, and it gives you raw material to turn one recording into a week of content. This guide covers the whole path from a finished episode file to published captions.

What a podcast transcript is actually for

The most common reason is repurposing. One 45-minute interview contains enough quotable material for a newsletter, a handful of social clips, and a written summary, and writing that by hand means listening to the tape three more times. A transcript turns the searching problem into a reading problem.

The second reason is reach. Listeners who are deaf or hard of hearing, people who are not native speakers of your language, and anyone who simply prefers reading all rely on the written version. Several podcast directories also treat transcripts as a quality signal, and some players will surface a transcript as a separate search result.

The third is searchable archives. Nobody scrolls back through forty episodes looking for the moment you explained pricing. With transcripts indexed, that sentence can be found.

Handling episodes longer than ten minutes

The ten-minute ceiling is a hard limit, not a soft warning, and it exists because the transcription model runs locally on limited hardware. Rather than compress or truncate your audio, split the file.

Split on structure, not on the clock. If your episode has segments, transcribe each segment as its own file. If it is a single conversation, split on a long pause or a clear change of subject. A transcript that reads in the right order afterwards is one you can publish without editing.

When you reassemble the parts, keep a small amount of overlap. Models occasionally drop the first few words of a file, so overlapping by a second or two on each join means no sentence is lost at the seam.

Two-speaker audio and music beds

A conversation between two hosts is a harder case than one speaker reading. The model does not label speakers, so if you need to know who said what, you will be matching the transcript against the audio yourself. Word mode helps here, because tighter timings make it easier to follow a line back to the recording.

Intro and outro music is a different problem. Music is not speech, so it produces either nothing or a scattering of mis-transcribed words. The fix is to trim the bed before upload rather than trying to clean it afterwards.

Remote recordings are worth checking first. Two people recorded on separate machines in different rooms arrive with noticeably different levels, and the quieter speaker is where most errors appear.

Publishing the transcript

For an audio show, put the transcript on your own site under a stable URL and link it from the episode description. Host it yourself rather than relying only on a platform page, since you want it indexed against your domain.

For a video show, the captions are the transcript. Upload the SRT to the platform and keep the file; the platform will also generate its own, and having yours means the timing is the timing you intended.

Whichever you choose, export early. Results are held for 60 minutes, and a finished episode transcript that expires while you are writing the show notes is an afternoon you will not get back.

Step by step

1

Split the episode into parts that fit the limit

Konthora accepts up to 10 minutes per file. Most podcast episodes are 30 to 90 minutes, so split yours at a natural break — a topic change, a sponsor read, or the end of a segment — rather than an arbitrary timestamp. Transcribe each part and keep the pieces in order.

2

Upload the part you want captioned

Drop the file into the workspace. If your show publishes video, upload the MP4 and the audio track is extracted automatically. No account is needed.

3

Pick sentence timestamps for show notes and highlights

Sentence mode gives each line a start and end time, which is what you want when you are pulling quotes or building chapter markers. Use word mode only if you need subtitle-accurate alignment.

4

Run the transcription and wait for the result

Processing happens on a CPU worker and takes roughly the length of the audio. Episode music beds and overlapping speech are the two things most likely to need a manual pass afterwards.

5

Export SRT or VTT if you publish video, TXT if you do not

Audio shows take a plain transcript. Video shows need SRT or VTT for the player or the platform. Export before the 60-minute retention window closes.

Open the transcription tool

Frequently Asked Questions

Can Konthora transcribe a full 60-minute episode in one go?
No. Media is capped at 10 minutes per file, so a full episode needs splitting into parts. Split at a topic change or a sponsor break rather than an arbitrary timestamp, overlap each join by a second so no sentence is lost, then export the parts in order.
Will the transcript tell me who said what?
No. Speaker labels are not produced, so for a two-host or interview episode you will match the text back to the audio yourself. Word-level timestamps make that job easier because each word has its own timing.
What happens to the music in my intro and outro?
Music is not speech, so it produces either nothing or a scattering of mis-transcribed words. Trim the intro and outro beds out of the file before uploading rather than cleaning up the transcript afterwards.
Is there a size or format limit on podcast episodes?
Up to 100 MB and 10 minutes per file. Audio inputs are MP3, WAV, M4A and AAC; MP4, WebM and MOV are accepted too, and the audio track is extracted automatically for video shows.