Transcribe the recording, mark the ten or so moments where someone said something specific, and turn each moment into one standalone piece: a text post, a clip, or both. Write captions for every clip, have a person approve the set against the original, and schedule it across two or three weeks. Expect roughly half a day.
The usual advice is right in outline: transcribe, find highlights, cut clips. What it leaves out is the part that decides whether the result is any good. Most recordings produce a handful of strong ideas and a lot of connective talk, and tools that clip automatically cannot always tell the difference.
This note is the workflow we would hand to someone doing it for the first time, with a map of what each moment can become and a time budget labelled for what it is: guidance, not a benchmark.
Why does one recording hold so much content?
An hour of people talking about their work contains more specific ideas than most teams write in a month.
In a podcast or a webinar, people answer questions they did not fully prepare for. They reach for examples, tell stories about mistakes, disagree with each other, and say the plain version of what the website says carefully. Those are exactly the things that work as LinkedIn posts: one idea, stated specifically, by a real person.
The mistake is to treat the recording itself as the content. Nobody scrolling a feed will watch an hour. They might read a sharp paragraph or watch forty seconds of someone saying one clear thing. So the job is not to shorten the recording. It is to find the moments that stand on their own, and give each one the format it suits.
That is also why fully automatic clipping disappoints. A tool can find moments where the speaker is animated or the words sound quotable. It is less reliable at knowing whether the point depends on something said five minutes earlier, whether the joke only works in context, or whether a guest would be comfortable seeing that sentence on its own.
What is the workflow from recording to schedule?
Six steps, in this order. The order is the point.
Copy this and run it the day after the recording, while the context is fresh.
- Transcript. Run the recording through a transcription tool (many recording and editing tools, such as Descript or Riverside, include one) and fix speaker names and obvious errors in product names. Keep the timestamps.
- Pull-quotes. Read the transcript, or ask a model to propose candidates and then read around each one yourself. Mark every moment that makes one specific point: a claim, a how-to, a mistake, a number the speaker can stand behind, a disagreement, an audience question. Note the timestamp and a one-line summary.
- Posts. Turn the strongest text moments into standalone posts. Each post gets its own opening line, the idea, the example from the recording, and a closing thought or question. Do not paste transcript chunks; spoken language reads badly as text.
- Clips. Cut the strongest spoken moments into short clips, starting on the first sentence that makes sense on its own. Cut the throat-clearing. Keep the speaker’s face if you have it.
- Captions. Burn in captions on every clip, check them against the audio, and write the post text that goes with each clip. Many people watch with sound off, and captions are also what makes the clip accessible.
- Approve and schedule. A person reviews the full set against the original recording, the speaker approves anything that quotes them, and the approved set is spread across two or three weeks on the calendar.
Steps two and six are where the quality comes from. A model can do much of the transcription, candidate-finding, drafting and captioning. It cannot be responsible for whether a quote is fair to the person who said it. That stays with a person who heard the whole thing.
What can each kind of moment become?
Pick the moment first, then the format that suits it.
Not every moment is a clip, and not every moment is a post. Some are best read, some are best heard, and some make good material for both. This map is a starting point for a typical one-hour conversation that yields ten posts and five clips.
| Kind of moment | Best format | How many from one hour | What to check |
|---|---|---|---|
| A strong, specific claim | Clip plus post | 2 clips, 2 posts | That the claim does not depend on earlier context |
| A step-by-step explanation | Text post or document post | 2 posts | That the steps are complete when read alone |
| A story about a mistake | Clip | 1 clip | That the speaker is comfortable with it standing alone |
| A number or result | Text post | 1 post | That the speaker can back it up and has cleared it for sharing |
| A disagreement between speakers | Clip | 1 clip | That both sides are represented fairly |
| An audience question | Text post answering it | 2 posts | That the answer is complete, not just the question |
| A memorable one-liner | Clip or quote image | 1 clip | That it is not funny only in context |
| The overall theme | Summary post linking the full recording | 1 post | That it gives a reason to watch the full recording |
| What you would add now | Follow-up post | 2 posts | That it adds something the recording did not say |
If a recording does not yield ten good moments, publish fewer. Five strong posts do more than ten padded ones, and a reader who meets three weak posts from the same webinar learns to scroll past the fourth.
How long does it take?
Budget about half a day for a one-hour recording, most of it spent reading and checking.
These are our working estimates for someone who has done it a few times, using ordinary tools. They are guidance to plan a week, not measured benchmarks, and a first attempt will take longer.
Rough time budget for one one-hour recording, as guidance.
- Transcript and clean-up: about 20 to 30 minutes, mostly fixing names.
- Reading and marking moments: about 45 to 60 minutes. This is the step not to rush.
- Drafting ten posts with a model and editing them: about 60 to 90 minutes.
- Cutting five clips: about 45 to 60 minutes.
- Captions and clip post text: about 30 minutes.
- Review, speaker approval and scheduling: about 30 minutes of your time, plus however long the speaker takes to reply.
The biggest time saving is not a faster tool. It is doing this every time you record, so the steps become routine, and keeping one template for posts and one caption style for clips, so nobody redesigns them each week.
What makes a short clip worth watching?
The first sentence has to make sense to someone who has never heard of you.
Start on the point, not the lead-up. If the speaker says “so, the thing about onboarding is, and we learned this the hard way, most users never see the second screen”, the clip starts at “most users never see the second screen”. The context goes in the caption or the post text.
Keep it to one idea. If the clip needs a second idea to land, it is two clips or it is a post. Crop for where it will be watched: vertical or square framing works better in a mobile feed than a wide shot with two small faces in it. Keep burned-in captions large enough to read on a phone, and check every word against the audio. Automatic captions get product names and people’s names wrong more often than anything else.
Before any clip goes in the queue, watch it next to the thirty seconds before and after it in the original. If the meaning shifts when you add the context back, the clip is not fair to the speaker, and it should not go out.
The first sentence has to make sense to someone who has never heard of you.
Who should approve, and how do you know it worked?
The person quoted has the final word on the quote.
Two approvals matter. The account owner approves the set as a whole: is this what we want to say, in our voice, on these dates. And every guest or colleague who is quoted approves the posts and clips that quote them. Send them the set in one message, with the dates, and ask for changes by a specific day. It is a courtesy that also catches mistakes.
To know if it worked, look past the reactions. Note which moments drew comments from the kind of people you want as buyers, which drove visits to the full recording, and whether anyone who engaged later booked a call. With a handful of posts, treat that as a pattern to test next time, not a result. Our note on how to automate LinkedIn posts with AI and still approve every one covers the read-back step in more detail.
A recording is not one piece of content. It is a list of things people said well. Find them, give each one its own shape, and let the person who said it have the last word.
Questions people ask next
How many LinkedIn posts can I get from one podcast or webinar?
A typical one-hour conversation can yield around ten posts and five short clips if it contains that many specific moments. If it does not, publish fewer. Strong posts do more than padded ones.
Can AI clip a podcast automatically?
AI can transcribe, suggest candidate moments, draft posts and generate captions. A person should still choose the moments and check each clip against its original context before it is published.
How long does it take to repurpose a webinar?
As rough guidance, about half a day for a one-hour recording once the routine is set, most of it spent reading, checking and editing rather than cutting.
Do I need permission to post clips of a guest?
Ask. Send every guest the posts and clips that quote them, with the dates, before they are scheduled. It is good manners and it catches quotes that do not stand alone.
Can Throughline turn our webinar into LinkedIn posts?
It drafts the posts and carousels in your brand from the moments you choose, for a person to approve, and reads your webinar attendee data to show who went on to book. Cutting the video clips stays with your editor.