How to Write Show Notes, Chapters and Social Posts from a Podcast

1. Start with existing content or a new recording

Open Tools → AI Writer, then Create drafts. Choose existing content with an available transcript, or upload a file or paste a supported content link. MP4 video and MP3/WAV audio use the same source checks as clip creation. Confirm that you own the material or have permission to process and adapt it.

The first step selects your source and output language. If you want writing in another language, enable Translate and choose the target there. That choice controls the writing language; changing the interface language later does not translate an existing draft. Without translation selected, the writing follows the source language.

Next opens the writing settings without uploading or using processing time. For a new source, the second step’s explicit upload-and-generate action starts uploading, transcription and the selected writing task. You can leave after the request has been accepted; writing continues after transcription finishes. Existing eligible content reuses its transcript. You do not need to create a video clip merely to write show notes.

2. Select the formats that serve your reader

AI Writer offers five formats: Show notes, Chapters, Newsletter, Short social post and LinkedIn post. Select the ones you intend to review. Show notes can introduce the episode and its useful points; chapters organize the recorded discussion; social posts introduce an idea to someone who has not listened yet. Newsletter includes a subject, preview text, body and call to action, and requires Lite or Standard.

For social posts, you can use one saved Clip as the source. Other formats use the whole processed source range. A clip about one decision should not become a summary of claims elsewhere in the episode. Add writing instructions that help the intended reader, such as the audience, tone or the point the post should explain.

Set the publishing identity carefully. Social writing defaults to third person; do not imply that you are the guest or that the guest endorses your product. Chapters depend on reliable source times. When adding or changing chapter times yourself, use actual points in the recording and keep them in increasing order.

3. Separate processing minutes from writing rounds

A new recording uses processing minutes for its selected processing range. Reusing an available transcript does not use those processing minutes again. Writing has a separate lifetime allowance for each source: 1 round without an account, 1 on Free, 3 on Lite and 10 on Standard, including the first generation. These rounds do not reset monthly. Languages and Clip selections share that source’s allowance.

Several formats submitted together use one round. Rewriting a single item also uses one round, so make ordinary wording changes manually when that is all you need. Manual edits, saving and copying do not consume a writing round. Compare the plan limits before committing to a sequence of different generations.

If some selected formats succeed, the round is used; use the original task’s recovery action to fill in failed items with the same settings without another round. If all writing fails, the round reservation is released. A writing failure does not undo successful transcription or refund its processing minutes. If transcription itself fails, recover source processing in My content, then return to the writing task.

4. Compare the candidate with the recording

Open each generated format and read its source references. On a wide screen the evidence appears beside the text; on a phone, open the source drawer. Use a reference to listen to the relevant passage when the original media is available. Check names, numbers, claims and qualifications before accepting the draft.

Generated writing can summarize or rephrase an idea. It still needs an editorial check: a guest describing one experience is not necessarily recommending that approach to everyone. Review any “I” or “we,” any promised outcome and any call to action. Add your own context where appropriate without presenting it as a quotation.

Adopt a candidate into the editor, make your changes, then save the document. New generations do not automatically replace a saved human-edited draft. If another window has saved a newer version, compare that version with your local changes before deciding what to keep. The source comparison also matters after changing a Clip’s range; older writing does not automatically become an accurate description of the new cut.

5. Download the version you have actually saved

Copying is available without an account; file downloads require registration. Download the saved writing as TXT, Markdown or ZIP. A ZIP can combine selected saved documents and eligible completed MP4 files. It does not render a new video merely because you included one in the selection.

If the chosen writing has unsaved edits, choose whether to save them first, download only the saved version or cancel. Open the downloaded file and check its title, language and contents before sending it. AI Writer does not send newsletters or publish social posts automatically. Prepare publication in your own email or social workflow.

Existing newsletters remain readable, copyable and downloadable after a downgrade, but generating or editing a Newsletter still requires the appropriate plan. For a visual companion to the post, use the separate cover and quote-card workflow.

6. Keep track of tasks, saved drafts and retention

Previous drafts shows your writing tasks, with source titles, formats and status. Open the task you need rather than assuming a newer successful result belongs to an earlier request. The task’s original candidate and the source’s current saved document may differ because you edited the latter afterward.

Writing shares the source’s retention deadline. Creating a new task or saving a document does not restart that clock. If the source is on day 20 of a 30-day retention period, its writing still expires with it on day 30. Download the saved documents you need while they remain available, and check an uncertain request’s status before trying to generate it again.