Turn a reference video into a shot list you can inspect.

1. Choose the output you actually need

Start with a video you own or are permitted to analyze. Decide what you want to take away before submitting it:

• Only the words: use a transcript workflow. It cannot describe a camera view or visual action that is never spoken.
• A reference breakdown: use Video Breakdown AI to inspect the opening, shot descriptions, structure and script together.
• An exact edit timeline: inspect the source in an editor. This guide does not promise frame-accurate cut boundaries, an EDL or a ready-to-import editing project.
• A plan for a new shoot: add your own approved script, people, equipment, locations and responsibilities after reviewing the reference.

The steps below focus on the second task: a reviewable record of an existing video, with structured data you can take into your own workflow.

2. Configure one reference video

Video Breakdown AI documents support for YouTube videos and Shorts, TikTok videos and other supported public video sources. Start with one direct video URL rather than expanding an entire channel while you are checking the output.

Configuration example, not an executed request:

{
  "urls": ["YOUR_SUPPORTED_VIDEO_URL"],
  "outputLanguage": "English",
  "translateQuotes": false,
  "includeTranscript": true,
  "maxVideoMinutes": 3,
  "accountSummary": false,
  "proxyMode": "datacenter"
}

Replace the placeholder with your permitted reference. The three-minute limit is a deliberate short-reference setting for this example: longer videos are skipped, not clipped into a three-minute excerpt. English sets the analysis language, while translateQuotes: false keeps quoted speech and on-screen text in their original language. Current input reference.

Before starting, review the current pricing. Video analysis, extra duration, Actor Start and optional residential traffic are different cost components. This example does not select residential mode, but it is not a free-run configuration or an all-in cost quote.

3. Keep the complete JSON before making a spreadsheet

Export the original dataset as JSON and keep it with the source URL and run identifier. The documented output includes video-level information and nested analysis fields. Decide which level each destination needs:

On a narrow screen, scroll the table sideways to read all columns.

Do not replace the JSON with a flattened summary sheet. A later reviewer may need a field the sheet never included. Keep unsuccessful rows out of creative findings, but retain them separately for diagnosis.

4. Make the shot review explicit

The current README's shortened shot example uses start, end, shot, scene, action, expression, on_screen_text and spoken inside shots. In that example, shot carries a framing label. Inspect your returned objects rather than assuming every field is filled or that the displayed table contains the whole object. Documented shot example.

Download the blank video-shot-review CSV. It is a proposed worksheet with headers only, not extracted footage or a claimed Actor result.

Use one row per shot entry you are reviewing:

1. Copy the source identity. Keep the video URL and your run/dataset reference on every row, so the row still makes sense if someone sorts or filters the sheet.
2. Keep the original position. Record the entry's index and its reported start and end. Avoid overwriting them when you later correct the timing.
3. Map the descriptive fields. Put the documented shot, scene, action, on_screen_text and spoken values into the corresponding draft columns. Leave missing information blank rather than inventing it.
4. Watch the referenced interval. Compare the description, text and dialogue with the video. Mark the review as unchecked, checked or needs correction.
5. Record corrections separately. Use reviewed start/end and reviewer-note columns. This preserves the difference between what the tool returned and what you verified.

For automation, this requires a separate flattening step over each successful video's shots array. The blank worksheet does not perform that transformation, and this page does not provide an already-tested shot-row exporter.

Check several things during review: an action may continue inside one camera view; a brief visual detail may be missed; a description may combine events; and quoted text can need correction. For precise editing, check boundaries against the source timeline rather than treating an AI description as an edit instruction.

5. Understand the existing n8n template's boundary

TubeText's existing n8n workflow is useful for a one-row-per-video research sheet. Its current formatter includes the hook, summary, structure, key moments, shot count, CTA, on-screen text and transcript, among other video-level fields.

It does not map the full shots array to the Google Sheet, and it does not create one spreadsheet row per shot. Preserve the original dataset if you need the detailed shot list, then add and validate your own shot-level destination. A populated shotCount cell is not a substitute for those descriptions.

Read the setup and failure-handling notes before importing. The template uses credentials you configure in n8n and a Google Sheet you select. Successful analyses and diagnostic rows follow different branches. A successful visual analysis can lack a transcript; missing speech alone should not be rewritten as a visual-analysis failure.

The template uses a synchronous Apify request. A timeout needs investigation of the original run and dataset before another attempt. Importing a JSON template does not connect accounts or prove that your live workflow is configured correctly.

6. Use the review for a decision

Once the shot rows are checked, ask a narrow question your team can act on: Where is the product first visible? Does the opening explain the problem before showing a demonstration? Which spoken line overlaps the reveal?

Keep each answer tied to a source interval. Then write a separate idea for your own work. For example, “try showing our product before the explanation” is a proposed creative choice. It is not evidence that the reference converted well or that the same choice will work for your audience.

Do not infer the original camera model, exact lens, production budget or campaign results from the breakdown. Do not republish another creator's footage, script or claims just because you have an analysis of it. The creative-research guide shows how to carry reviewed observations into a brief without treating an AI interpretation as performance evidence.

Common questions

Can this generate a shot list from an existing video?

Video Breakdown AI returns a documented shots analysis that you can inspect and organize. Treat it as a draft description of the reference. It is not a guarantee that every visual event or edit has been captured.

Is the output a production-ready storyboard?

Not by itself. This workflow gives you a structured reference and a review process. Shooting decisions, illustrations, rights checks and production logistics remain separate work.

Can I use Google Sheets without losing the detail?

Yes, if you deliberately retain the original JSON and build a shot-level table from it. The existing n8n template is a video-level comparison sheet; its standard formatter does not preserve the full shot array.

Start with one reference

Open Video Breakdown AI on Apify, review its inputs and costs, and inspect one permitted video before building a batch workflow. Keep the source, the raw analysis and your corrections together.

Only need spoken content? Start with YouTube captions and optional AI transcription instead.

TubeText Labs maintains Video Breakdown AI and the linked example repository. This page was prepared with AI assistance from the linked documentation and workflow source. The configuration and blank worksheet are not a new live test, frame-accuracy benchmark or promise of campaign results. Product execution and billing happen in your Apify account. About TubeText.