Podcast clip service: sell short-form from a back catalogue

Most podcasts publish an episode a week and clip almost none of it. The back catalogue — two hundred episodes, four hundred hours — sits there containing every good line the host has ever said, and nobody can find them, because finding them means listening again. Text solves the finding problem, and the finding problem is the whole job.

This playbook is the service built on that: transcribe a channel’s episodes, locate the moments worth clipping using the timestamps, and hand the client a specification an editor can cut from without opening the audio. The input cost is half a cent per episode. The service sells for the price of a freelance editor’s afternoon.

What you are actually selling

Editorial judgement with the search cost removed. The client is not paying for transcripts and they are not paying for video editing — they are paying for someone to have decided which forty seconds of a ninety-minute conversation is worth posting.

The deliverable is a clip spec, one row per clip:

• Start and end timestamp, to the second, with a link that opens the video at that moment
• The verbatim quote, so the client can approve it without watching
• A hook line for the first two seconds
• Suggested caption and title text
• One line on why this clip works

An editor takes that and cuts. A social manager takes it and schedules. Neither has to listen to the episode. That is the thing being bought.

Sell it as a monthly retainer over a fixed episode count, not as a one-off catalogue project. Catalogue projects end. Retainers are the business.

The numbers

One month of service for a weekly podcast: four new episodes, eight clip candidates each, thirty-two clips delivered.

The 99.7% figure is the one that gets quoted and it is the less useful of the two. Data is not your cost structure. Your cost structure is the four hours you spend reading candidates and rejecting the ones the model liked and you do not.

Cost the hours honestly at the rate you would otherwise bill, and the business is a 60% margin service business with an unusually cheap input — which is a good business, and a different claim from “it costs nothing”.

The onboarding run is separate. Two hundred back-catalogue episodes is $1.00 of transcripts. Charge for the catalogue sweep as a one-off setup fee; it is the strongest pitch you have and it costs you a dollar.

How it works

1. Take the client’s channel URL and run the back catalogue once. Save the raw output to disk and do every later pass on the saved file. Re-running to fix a formatting mistake is wasteful and slow.
2. Keep timestamps on. Without them, you have quotes you cannot locate, and a clip spec that cannot be cut from is not a deliverable.
3. Segment each episode into topic blocks. Ten to fifteen per ninety minutes. A block is a coherent stretch on one subject, which is what a clip has to be.
4. Score the blocks for clippability. Self-contained (needs no prior context), a claim rather than a hedge, a clean start and end, and 30–90 seconds long. Discard anything that opens with “as I was saying”.
5. Draft the spec for the top eight. Quote verbatim from the transcript — never paraphrase into the quote field, because the client will check it against the audio and a mismatch destroys the deliverable.
6. Watch the eight before you send them. This is the step that is tempting to skip and the one that catches the clip where the host is being sarcastic, or where the good line is answering a question from a guest you did not include.
7. Deliver as a spreadsheet plus a shared doc. Editors want a table. Clients want prose. Send both.
8. Track which clips got made and which performed. After two months this is your scoring signal and your renewal argument.

Where it gets hard

Finding clients is harder than delivering the work. Podcasters with budget already have an editor, and podcasters without budget are the ones with the time to clip their own episodes.

The buyer you want is a show with sponsorship revenue, no in-house social person, and a back catalogue over fifty episodes. That is a narrow band. Expect to pitch cold, and expect the first ten pitches to go nowhere.

Lead with a free sample. Run one episode of the target’s own show, deliver three clip specs unasked, and let the specificity do the pitching. The sample costs you half a cent and an hour, and it converts far better than a description of the service.

The competition is real. Agencies sell this bundled with the editing, which is what most clients would rather buy — one invoice, finished videos, no coordination. Selling specs only means selling to clients who already have an editor. Either accept that constraint or add editing and become a different, more labour-intensive business.

Taste does not transfer. A model will nominate the passages that look quotable, and about half of them are hedged, context-dependent, or dull in the room. Reading and rejecting is most of the four hours, and there is no version of this where you stop doing it.

The boring parts are the business: invoicing, chasing approvals, clients who go quiet for three weeks and then want the whole month delivered on a Friday. Price the retainer so that a slow month still pays.

Limitations

• Episodes without captions return nothing. If the uploader disabled captions and no automatic track exists, there is no transcript to retrieve, and no tool can produce one from published data. Audio-only podcasts that are not on YouTube are out of scope entirely.
• The transcript is not the audio. Sarcasm, timing, laughter, a pause that makes the line — none of it is in the text. A passage that reads brilliantly can be flat, and you will only find out by listening. Always watch before delivering.
• No speaker separation. A two-host conversation comes back as one stream. Attributing a quote to the wrong person in a spec is the mistake that loses a client, so verify attribution on anything you send.
• Automatic captions get names wrong. Proper nouns, company names and technical terms are where errors concentrate, and those are disproportionately the words inside a quotable claim. Check every quote against the audio.
• Timestamps drift on long episodes. They are accurate enough to locate a moment, not to frame a cut. Treat the timestamp as a pointer for the editor, not as an edit decision list.
• The content is the client’s. You are working on their material with their permission, which is the whole basis of this being legitimate. Transcribing someone else’s podcast to sell clips of it to a third party is a copyright problem, not a business.
• Private and unlisted videos return nothing. If it does not open in a logged-out browser, it cannot be extracted.
• You cannot promise performance. Clips are a numbers game with a wide variance. Sell volume, judgement and consistency; do not sell view counts, because you do not control the account, the thumbnail, the posting time or the algorithm.

FAQ

How many clips are actually in one episode?

Six to ten genuinely good ones in ninety minutes of conversation, in our experience of reading them. You can nominate thirty candidates and most of them will be mediocre. Promising eight per episode is deliverable; promising twenty is how you end up shipping padding.

Do I need to edit video to sell this?

No, and staying out of editing is what keeps the margin. The trade-off is that you can only sell to clients who already have an editor, which is a smaller market. Adding editing roughly triples your hours per episode and changes what you can charge accordingly.

What do I charge for the back-catalogue sweep?

Price it on the work, not the input. Two hundred episodes is $1.00 of transcripts and perhaps twelve hours of reading if you are selective about which episodes to mine. A flat setup fee of $1,500–3,000 is defensible, and it is usually the easiest thing to sell because the client can see the catalogue sitting unused.

Does this work for podcasts not on YouTube?

Not with this approach. This works from published video with captions. An audio-only show distributed by RSS needs speech-to-text instead, which costs meaningfully more per hour and changes the arithmetic on the sweep — though the rest of the service is identical.

What stops the client doing it themselves?

Nothing, and some will. The service sells on the client not wanting to spend four hours a month reading their own transcripts. Be honest about that in the pitch; pretending there is a secret method is a worse position than pretending nothing, because the client will find the tooling and feel misled.

How do I price the first client?

Below your target rate, in exchange for a testimonial and permission to show the work. One reference client with visible published clips is worth more than the discount, and the second sale is far easier than the first.