How to Turn a Podcast Clip Into Three Quote Cards

Anyone sitting on a good two-minute podcast clip can turn it into three quote cards in 20 minutes: three sentences that stand alone, square 1:1 frames, on-screen text under 12 words. Use ChatGPT or Claude to pick the lines. Use any editor or a CapCut-class tool to build the cards. Do not recap the episode.

A quote card is a still people can read with the sound off. If the line only works after 40 seconds of setup, it is not a card.

Keep type readable the way you keep text readable on an AI poster: few words, high contrast, no script font. If you also write the post caption, keep it human with captions that do not sound like ads.

Make three quote cards in 20 minutes

1. Pull the clip, not the episode

Export or mark one clip, two minutes or less. One topic. If the file is the whole show, you are about to recap. Stop and cut first.

Transcribe just that clip. Paste the transcript into ChatGPT or Claude. Instruct: "Do not summarize the episode. List every sentence that can stand alone with no setup. If a line needs the question before it, drop it. Cap the list at eight. I will pick three."

The model will offer a theme. Theme is recap. You want sentences. A podcast intro under 12 seconds is a door for the audio. These cards are a door for the still. Do not put the sting on the card.

2. Pick three lines that survive silence

Read the candidate list with the sound off in your head. If you need the guest's name or the last anecdote to care, cut it.

Choose three. Not six. Hard tests: under 12 words after a tight cut, one claim per card, no "that's a great question."

Instruct: "Cut each chosen line to 12 words or fewer without changing the meaning. If you cannot, mark FAIL. Do not invent a punchier version I did not say." FAIL means you picked a ramble. Go back to the list.

3. Build square 1:1 cards, not a timeline

Open the editor. Three frames. 1:1. Same background family for all three: one still from the clip, or one color field, or one blurred freeze. Do not switch palettes mid-pack.

Place the quote in one block. Under 12 words. High contrast. Leave margin. Do not put the words over teeth or a busy pattern.

Optional caption burn: same words as the quote, same order, burned for sound-off autoplay. Not a paraphrase. Not a second slogan.

Banned on the card: the episode number as a headline, a five-line bio, a URL in 8-point type, a fake pull quote the guest did not say.

4. One human listen against the waveform

Play the clip. Read each card out loud at the moment the line is spoken.

Check three traps:

• A cut that changed the claim.
• A name on the card the clip never said.
• Type smaller than a thumb can read.

Fix the words yourself. Do not send the card back to ChatGPT for a "punchier" rewrite. Punchy is how a quote becomes an ad.

If a card still needs a footnote, kill the card. Pick another sentence from the list of eight.

5. Export three stills. Do not add a recap slide

Export three images. Same size. Same naming pattern: show-slug-01, 02, 03.

You may overlay the card on a 15-second clip of the line. That is still one quote. If you start cutting B-roll under the type, you left this skill. Save the 12-word cap and the 1:1 size. Next clip, run the same loop.

Reusable skill card

Trigger: You have a short podcast clip and you are about to post a recap carousel or a 40-word screenshot.

Input: One clip, two minutes or less, its transcript, a request for stand-alone sentences only, a 12-word cap.

Output: Three 1:1 quote cards. On-screen text under 12 words. Optional burned captions that match the quote. No recap slide.

Stop: After three cards and one listen against the waveform. No sixth card. No punchier rewrite. No invented line.

Takeaways

• Three cards. One clip. Not the episode.
• A line that needs setup is not a quote card. Drop it.
• Under 12 words on screen. Square 1:1. Same pack, same look.
• ChatGPT or Claude picks candidates. You hear the clip and veto.
• Caption burn copies the quote. It does not paraphrase it.