vpv.

← Notes·2026-07-14·studio / video

Vanpelt Studio: checking your footage against the plan before you edit.

Long-form videos stall between the shoot and the edit. How to check footage against a written plan the same evening, and the $29 app I built to run that check.

By updated 7 min read

The most expensive moment in a long-form video is not the edit. It is the day, deep into the edit, when you realise a shot you needed was never filmed. The fix is a loop you run on the evening of the shoot: write the plan as a list, check what you actually recorded against it, and film the gaps while a pickup (a short shot filmed after the main shoot) is still a twenty-minute job.

You can run that loop by hand with a document and a free transcription tool, and this note shows how. Vanpelt Studio is the desktop app I built to run it for me: $29 a month with the AI included, on macOS and Windows, with footage that never leaves your machine. Every video on this site is planned with it.

The note is for solo creators and small teams making videos of eight minutes and up, where the problem is structure rather than cutting.

The plan-check-pickup loop on one card

  1. 01

    Write the plan as a list

    Beats, then the shots and lines each beat needs. Output: a numbered list you can tick.

    before the shoot
  2. 02

    Transcribe and index the footage

    Run every clip through speech-to-text so you can search it like a document. Output: one text file per clip.

    during dinner
  3. 03

    Grade each line of the plan

    Got it, partial, or missing, with the clip name next to each. Output: a list of holes.

    30 min
  4. 04

    Shoot the pickups

    The same evening or the next morning, while the location, light and outfit still match. Output: the holes closed.

    20 min each
  5. 05

    Edit only when the list is complete

    Output: an edit that is about cutting, not about discovering what you have.

    the edit
The plan-check-pickup loop on one card. Times assume a ten-minute video and are approximate; the loop matters more than the clock.

Why long-form stalls between the shoot and the edit

Videos of eight, fifteen or thirty minutes are a structure problem, and structure is where a solo creator has nobody to lean on. The time does not go into cutting. It goes into deciding what the video is, watching hours of footage back to remember what is in it, and finding the missing shot when it is too late to get it cheaply.

Film crews have a name for the fix. A pick-up is a small, relatively minor shot filmed after the fact to add to footage already shot. On a feature, the producer weighs pickups against the cost of reuniting cast and crew. For a solo creator that cost is low on the evening of the shoot and high two weeks later, when the light, the location and sometimes the haircut have changed.

One hypothetical runs through every step: say you are making a ten-minute video about building a standing desk.

Step 1: write the plan as a list you can tick

A plan you can check against has to be specific enough to fail. "Talk about materials" cannot fail. "Hold up the receipt and say the total" can. The format that works is two columns: what is said on one side, what is seen on the other. That is the A/V script taught in production courses, and the right-hand column doubles as the shot list.

For the desk video, the list might read: a cold open of the wobbling old desk; the receipt and the total; cutting the frame; the moment the legs do not fit; the fix; a week of use, logged; the verdict said to camera. Seven lines, each one something you either filmed or did not.

The beat sheet guide walks through writing that list from scratch, and the script process turns it into the two-column version.

Step 2: turn the footage into text

Watching every clip back is where days go. Transcription fixes that. Once every clip is text, "did I ever say the total?" is a search, not an afternoon.

You do not need a paid tool for this step. OpenAI's Whisper is an open-source speech recognition model, released under the MIT licence, trained on 680,000 hours of audio, and it runs on your own computer from the command line. It needs ffmpeg, the free cross-platform tool for recording and converting audio and video. If you edit in DaVinci Resolve, version 21's IntelliScript goes further for scripted material: it compares the script text to the transcribed audio and assembles a cut.

Step 3: grade every line of the plan

With the transcripts open, mark each line: got it, partial, or missing. Write the clip name next to every got it, because that is your edit map. Partial is the dangerous grade: the line was said, but walking out of frame, or the shot exists but the receipt is unreadable.

For the desk video, the check might find two holes. The receipt shot is missing, and the verdict is partial because it was said while packing up, half out of focus. Both are pickups you can film in twenty minutes the same evening. Found on day twelve, the verdict alone could mean rebuilding the ending.

Footage against plan: the evening check

  • The plan is a numbered list, one shot or line per row
  • Every clip from the card is transcribed and saved as text
  • Each row is graded got it, partial or missing
  • Every got it row has a clip name next to it
  • Every partial row says what is wrong: framing, focus, audio, delivery
  • Pickups are listed with location, light and outfit to match
  • Pickups are filmed before the location changes
  • The edit starts only when no row is marked missing
A check you can run by hand with a text editor and a transcription tool. It is a coverage check, not a judgement of whether the story works.

The turn: the check cannot tell you the plan is wrong

A coverage check only catches what you wrote down. If the plan is wrong, every row can be green and the video still fails. My own two-minute vlog is an example: every line I needed was filmed, but the sentence the video exists for lands at 0:26, after the opening backstory. No footage check would flag that. It is a structure decision, and it belongs to the plan and the edit.

The product page says the same thing in its own terms: if your videos are already tightly planned and your footage is already organised, you will save less. The loop is worth the most when the mess is the problem. If you shoot to a tight script every time, a transcription tool and the checklist above may be all you need.

What the app does, role by role

The app is organised around the jobs a production team would do on a video, because the jobs are what a solo creator is missing.

Before the shoot

Decide what the video is.

  • Strategist: finds videos your audience already watches, then reads how each of yours did
  • Writer: turns the idea into a hook, a script and a shot list
  • Packaging: drafts titles and thumbnails and pressure-tests them

After the shoot

Find out what you actually got.

  • Assistant editor: transcribes and indexes every clip on your machine
  • Story editor: checks the footage against the plan and names what is missing
The five roles in Vanpelt Studio and which part of the process each one owns, as listed on the product site in November 2026. A map of jobs, not a feature list.

The centre of it is steps 2 and 3 of the loop above: your list, graded against what the memory card holds, the same evening. Every cut still belongs to you. It is not an AI editor, and it does not touch your timeline.

Local-first, and why that is a cost decision

Your footage never leaves your machine. Whisper transcribes on device, ffmpeg handles the media locally, and only derived text (a transcript line, a beat, a shot description) is sent to a model for reasoning. You choose the model explicitly: your existing Claude or ChatGPT subscription, your own Anthropic or OpenAI key, ten open models through OpenRouter, or the AI bundled with Studio.

That constraint is not a privacy slogan. It is what makes the assistant editor affordable. Uploading a full memory card to index it is slow and expensive; indexing it locally is neither.

What it costs, as of November 2026

PlanPriceAINotes
Studio$29 a monthIncluded, sized for steady weekly outputNothing to set up; cancel anytime
Bring your own AI, monthly$12 a monthYour Claude or ChatGPT subscription, or an API key7 days free
Bring your own AI, annual$120 a yearSame as monthlyTwo months free
Founding lifetime$89 onceBring your own50 seats, with an onboarding call

Prices are from the pricing page on 19 November 2026 and may change.

What came out of the same loop, free

Two things came out of making the channel's own videos. The free creator course is 11 modules and 56 lessons covering thumbnails, titles, ideas, story and scripting, with every lesson open. The free LUT pack is the set of grades from my own films, as .cube files licensed for monetised work: Retro Surf from the surf movie, and Horror Green and Horror Blue from the horror short. The LUT pack collected 300 email sign-ups in under a week.

If the loop breaks, it is one of these

  • Every row is green but the edit still stalls. Cause: the plan was a topic list, not shots. Fix: rewrite each row as something you can point a camera at.
  • The check takes longer than the shoot. Cause: watching instead of searching. Fix: transcribe first, then search the text.
  • Pickups do not match. Cause: filmed days later. Fix: check the same evening and note light and outfit for each pickup.
  • Partial rows pile up. Cause: lines delivered while moving or off-axis. Fix: on the day, repeat key lines once, still, to camera.

The whole loop on one line

Write a list that can fail, turn the footage into text, grade every row, and shoot the holes before the location changes. Do it by hand, or let the app do the watching. Either way, the twenty-minute pickup beats the lost week.

Client work runs through the same loop: a launch film on the video side of the studio is planned with the same list and checked the same way. The full build story is in the Vanpelt Studio case study.

Sources and further reading

Want a video like this made for your startup?

Video by the studio →Start a conversation →