How to Turn a Long YouTube Transcript Into a Shorts Cut List With NotebookLM

The problem and who this is for

The slow part of repurposing a long YouTube video is usually deciding what to cut, not cutting it. This workflow is for creators, editors, agency teams, and channel managers who already have a finished or nearly finished long-form video. The goal is to get to an editor-ready cut list with the best clip candidates, rough timestamps, hook ideas, and simple notes for captions or on-screen text with the fewest steps possible.

Prerequisites

Access to NotebookLM
A source file or capture workflow that matches the article
A place to save the final output, such as a Google Doc, notes app, spreadsheet, or editor brief
An editing or publishing workflow for your final short-form asset

How to capture or gather the source material

If the video is already published, grab the YouTube URL and check whether the transcript is usable. If captions are messy, export or copy the transcript and clean obvious errors before you upload it anywhere.
If the video is not published yet, export a transcript from your editing or transcription tool. Save it as a Google Doc, DOCX, TXT, or PDF. A simple text file is fine.
Keep speaker labels if there is an interview or multi-person discussion. They help the model understand who is saying what.
If you can, add rough timestamps every 30 to 60 seconds or at each chapter break. The timestamps do not need to be perfect. They just need to make it easy to find the passage later in your editor.
Remove obvious filler that will never become a clip, such as repeated sound checks, long sponsor reads, or dead air markers. Do not over-edit the transcript. Leave the surrounding context intact.

Step-by-step workflow

Create a new notebook in NotebookLM and upload the cleaned transcript. If you also have chapter notes, a show outline, or a content brief, add those as separate sources.
Ask NotebookLM for a wide first pass. Tell it to find every passage that could stand alone as a short clip, then group results by clip type such as strong opinion, practical tip, surprising fact, mistake to avoid, or punchy one-liner.
Run a second pass with stricter rules. Ask for only the best candidates that make sense without heavy setup, sound good when read aloud, and can fit a short-form format after trimming.
Have NotebookLM turn the shortlist into a cut list with one row per clip: working title, source citation, rough timestamp, why the clip works, first line on screen, and a suggested call to action.
Open the actual video in your editor and spot-check every candidate against the source. Keep only the lines that still land when you hear the delivery, pacing, and tone.
Move the final list into your editing workflow. A plain Google Doc or sheet is enough. The point is to start the edit with a ranked list instead of hunting through the timeline.

Tool-specific instructions

Primary tool: NotebookLM

NotebookLM is the best fit here because the job starts from source material, not a blank prompt. You want grounded clip ideas tied back to real source passages.
Upload the transcript as the main source. Add supporting context only if it helps choose better clips, such as a content brief or chapter outline.
Use the inline citations to jump back to the passage before you commit to any clip. This is the fastest way to avoid attractive but weak picks.
Ask for ranking criteria that matter for clips: standalone value, short setup time, emotional pull, clarity out of context, and rewatch potential.
Do not ask NotebookLM to write a final edited script first. Ask it to find and rank raw source moments. That is where it is most useful in this workflow.

Alternative: ChatGPT

Use ChatGPT if you already have a clean transcript and want faster rewriting after you have identified a few clip candidates.
Upload the transcript or paste only the relevant section instead of the entire show if the video is very long.
Ask it to turn selected passages into trimmed clip scripts, hook lines, caption options, and alternative opening frames.

Alternative: Gemini

Use Gemini if your transcript lives in Google Docs or Drive and you want a fast file-based analysis path inside the Google stack.
Gemini is also useful if you want to upload a supporting screenshot of your chapter map or content brief beside the transcript.
Once you have a short list, move into your editor and verify every candidate against the real audio.

Copy and paste prompts

NotebookLM first-pass clip finder

{
  "role": "You are a source-grounded short-form clip strategist.",
  "goal": "Find strong short-form clip candidates from my uploaded transcript.",
  "instructions": [
    "Use only the uploaded sources.",
    "Do not invent scenes, claims, or timestamps.",
    "Find passages that could work as standalone short clips with minimal setup.",
    "Group results by clip type.",
    "Prioritize moments with a strong opening line in the first sentence."
  ],
  "output_format": {
    "sections": [
      "clip_type",
      "candidate_quote_or_summary",
      "source_citation",
      "rough_timestamp_if_available",
      "why_it_could_work_as_a_short"
    ]
  }
}

NotebookLM final cut-list prompt

{
  "role": "You are turning source-grounded clip candidates into an editor-ready shortlist.",
  "goal": "Create a ranked short-form cut list from the strongest transcript moments.",
  "rules": [
    "Use only cited source material.",
    "Keep only candidates that can make sense without long setup.",
    "Prefer clips that can be trimmed to roughly 20 to 45 seconds.",
    "Avoid duplicate ideas that hit the same angle."
  ],
  "output_fields": [
    "rank",
    "working_clip_title",
    "source_citation",
    "rough_start_and_end_if_present",
    "hook_type",
    "why_this_clip_should_be_cut",
    "first_line_on_screen",
    "caption_angle",
    "light_editing_note"
  ]
}

Quality checks

Every suggested clip points back to a real source passage.
The first two seconds are strong enough to stop a scroll.
The clip still makes sense if a viewer never watched the long video.
The list includes different hook types instead of 10 versions of the same insight.
The final shortlist is short enough to edit this week. Five to ten strong options is usually better than 30 weak ones.

Common failure modes and fixes

The transcript is accurate but hard to navigate: Add rough timestamps, chapter markers, or section headers before upload. Even loose markers make the shortlist more useful.
The clips sound good in text but weak in delivery: Spot-check the source audio before editing. Delivery beats wording more often than people think.
Every candidate needs too much setup: Ask NotebookLM to filter for moments that start with the answer, the opinion, or the surprising detail first.
You end up with only educational clips: Ask for a balanced mix of practical tips, contrarian takes, mistakes, personal moments, and quick wins.

Sources Checked

NotebookLM Help: Learn about NotebookLM: https://support.google.com/notebooklm/answer/16164461?co=GENIE.Platform%3DDesktop&hl=en (accessed 2026-03-25)
NotebookLM product site: https://notebooklm.google/ (accessed 2026-03-25)
NotebookLM FAQ: https://support.google.com/notebooklm/answer/16269187?hl=en (accessed 2026-03-25)
OpenAI Help: File Uploads FAQ: https://help.openai.com/en/articles/8555545-file-uploads-faq (accessed 2026-03-25)
OpenAI Help: ChatGPT Image Inputs FAQ: https://help.openai.com/en/articles/8400551-chatgpt-image-inputs-faq (accessed 2026-03-25)
Gemini Help: Upload and analyze files in Gemini Apps on Android: https://support.google.com/gemini/answer/14903178?co=GENIE.Platform%3DAndroid&hl=en (accessed 2026-03-25)
Gemini Help: Upload and analyse files in Gemini Apps on Computer: https://support.google.com/gemini/answer/14903178?hl=en-NA&visit_id=639100585389477783-308453270&p=code_upload&rd=1 (accessed 2026-03-25)

Quarterly Refresh Flag

Review by 2026-06-23. Re-check the current tool interface, upload behavior, supported source types, and any changes that affect this workflow before republishing or refreshing the article.