Use case · Creators
Drop in one episode and ship the whole content pack: titles and show notes, a newsletter edition, five social posts, and quote cards rendered as square images. This is the work podcasters rent per-seat repurposing tools for, or skip entirely — here it’s one canvas that already knows your show’s voice.
An episode’s value doesn’t stop at the RSS feed: the episode page, the email list, and the social feeds each need their own version of it, every week. That recurring rewrite is exactly what this graph automates — transcribe once, fan out to every format.
How it works
One transcription, four writer branches. Everything downstream of the transcript runs in parallel.
- Feed in the episode. The recording, a line of show context (name, host, voice, call to action), and how many quote cards you want.
- Transcribe once. A speech-recognition model turns the episode into a transcript that every branch reads from.
- Write the episode page. An agent drafts three title options, a summary, chapter-style show notes, a resource list, and the closing CTA.
- Draft the newsletter. A second agent writes the email edition in the host’s voice — subject-line options and a body that teases instead of summarizes.
- Generate the posts. A list generator writes five standalone social posts in mixed formats: a bold take, a question, a mini-story, a stat, a lesson.
- Render the quote cards. Another list generator pulls the most quotable verbatim lines, and each one becomes a square typographic card via a text-to-image model.
One recording, a week of content
Everything reads from the same transcript, so the pack is consistent: the newsletter teases what the show notes explain, and the quote cards say it in the host’s own words. Run it as the last step of every edit session and publishing day becomes an upload, not a writing shift.
Make it yours
- Tune each voice separately. Show notes, newsletter, and posts each have their own prompt and system prompt — make the newsletter personal and the posts punchy without compromise.
- Swap the transcription model. Whisper variants, local or hosted — the branch prompts don’t change.
- Restyle the cards. The quote-card prompt controls typography, palette, and format. Match your cover art, or go 9:16 for stories.
- Add branches. A YouTube description, a LinkedIn essay, a blog post — each is one more Prompt → Agent pair reading the same transcript.
Models in this workflow
The template ships with no models selected — pick one per role when you open it. Called with your own keys; the bill comes from the provider.
| Role | Node | Works well with |
|---|---|---|
| Transcription | Automatic Speech Recognition | Whisper (hosted or local) |
| Show notes, newsletter | Agent | Any strong language model |
| Posts, quotes | List Generator | Any language model |
| Quote cards | Text To Image | A model with reliable text rendering (Nano Banana, GPT Image) |
See Models & Providers to set up keys.
Next steps
- Open the template: Templates → Podcast Repurposing Studio in the dashboard
- Ad Creative Factory — the same fan-out, pointed at paid social
- Transcribe Audio — the transcription building block on its own
- All use cases