What This Is
This prompt is designed to help you turn raw podcast transcripts into clean, readable text — and then convert that text into a structured article draft ready for publication. It performs a two-stage workflow:
- Cleans the transcript by removing filler words and normalizing the dialogue while preserving tone and meaning.
- Builds an article draft that summarizes the conversation, captures key insights, and presents them in a clear, action-oriented narrative.
This is the same workflow I use for preparing The Union Podcast content before editing articles for krista.ai.
How to Use It
- Copy and paste the entire prompt below into ChatGPT.
- Paste your raw transcript beneath it — you can include timestamps or speaker labels if you have them.
- The model will first output a cleaned transcript, then automatically generate a draft article based on your cleaned text.
- If you only need the transcript, stop after Stage 1. If you want an article, let it continue to Stage 2.
You can reuse this same prompt for any podcast, interview, webinar, or fireside chat to quickly produce consistent, high-quality written content.
Optimized For
This prompt is optimized for GPT-5, which delivers improved natural-language understanding, tone preservation, and structural editing over previous models. It leverages GPT-5’s advanced speech-to-text reasoning to recognize hedging, pauses, and natural conversation markers — making the cleaned transcript sound human while removing clutter.
If you’re using GPT-5 or GPT-5 Turbo in ChatGPT or the API, no additional settings are required — just paste the prompt and your transcript.
Podcast Transcript → Clean Transcript → Article Draft Prompt
Stage 1 – Clean the Transcript
Goal: Clean and prepare a podcast transcript for article creation.
Instructions:
I would like you to remove all filler words and phrases such as “um,” “uh,” “like,” “you know,” “so,” “actually,” “basically,” “seriously,” “literally,” “I mean,” “right,” “okay,” “yeah,” “kind of,” “sort of,” and similar expressions that do not add meaning.
- Remove only when doing so does not alter tone, pacing, or meaning.
- Keep every idea and line of dialogue intact — do not summarize or omit.
- Maintain consistent speaker labels (e.g., “Scott:”, “John:”) and line breaks for readability.
- Preserve a natural conversational rhythm that reflects how people actually talk.
- When a filler contributes to humor, contrast, or authenticity, keep it.
- The output must read naturally and be ready for article conversion.
- Do not use em-dashes (—). Replace them with commas or periods where appropriate to maintain clarity and tone.
Deliverable:
A clean, natural transcript that reads smoothly and can be directly converted into an article.
Stage 2 – Create the Article Draft
Task: Write an article based on the cleaned transcript.
- Write in Scott King’s tone: authoritative, direct, and action-oriented. Avoid passive tense and appositive phrases.
- Preserve the core ideas and insights from the conversation while tightening the flow into a clear narrative.
- Structure the article with:
- Hook / Context: Introduce the main theme or problem discussed in the episode.
- Key Insights: Present the main takeaways from each speaker, weaving them into a cohesive argument.
- Practical Value: Translate the discussion into actionable insight or a business implication.
- Conclusion / Forward View: End with what’s next — the bigger idea, a question, or a takeaway statement.
- Retain quotes from John and Scott where they add emphasis or authenticity.
- Keep the article under 900 words unless more length is needed for clarity.
Deliverables:
- A cleaned transcript.
- A polished article draft suitable for publication on krista.ai or LinkedIn.
How to Use
- Paste the raw transcript beneath this prompt.
- The model will first output the cleaned transcript (Stage 1).
- Then it will automatically generate the article draft (Stage 2).
- You can stop after Stage 1 if you only need transcript cleanup, or continue to Stage 2 when drafting content.