Winery Reels AI Prompt: Turn Phone Footage Into Vertical Video

You turn phone footage into Reels by writing a plain text log of what's in each clip, pasting that log into Claude or ChatGPT with a prompt that asks for a shot-by-shot edit plan, then cutting to that plan in CapCut. The AI never touches your video files. It reads your description of the footage and tells you exactly which clips to keep, how long each runs, what text goes on screen, and what the caption should say on each platform.

That matters because the footage is already sitting on somebody's phone. Most tasting rooms, taprooms, and stillhouses have hours of usable video that never gets posted, because deciding what to cut takes longer than shooting it. This gets you from 20 minutes of raw clips to three finished vertical videos in an afternoon, without hiring anyone.

What this replaces

The normal path is either paying a freelancer a few hundred dollars per video or having your tasting room manager stare at a timeline for two hours and give up. The edit plan removes the hardest part, which is the decision of what the video is actually about and where it starts.

You still do the editing. But you do it with a script in front of you instead of guessing, and the whole thing takes about as long as a staff meeting.

What you need before you start

1. The footage. Twenty minutes is plenty. First pour of a new release, a barrel sample, the crush pad at 6 am, a staff member explaining why they like a wine. Unscripted is fine.
2. A footage log. Scrub through once and write one line per clip: file name, rough timestamp, what happens. For example: "IMG_4412, 0:14, Maria pours the Albariño and says it smells like the coast." This takes about five minutes and it's the only prep that matters.
3. An AI account. Claude or ChatGPT. The free tier handles the writing.
4. CapCut, on desktop or phone, for the actual cut.

A brewery logs the mash-in and the canning line. A distillery logs the spirit coming off the still. A wine shop logs a staff pick pulled off the shelf and talked about for 20 seconds. A distributor rep logs a buyer tasting something at the bar, with permission.

What it costs

Claude's free plan is $0 and covers this job. Claude Pro is $17 a month with the annual subscription discount ($200 billed up front), or $20 if billed monthly, per Anthropic's pricing page read on 2026-11-08. ChatGPT Plus is $20 a month, per OpenAI's help page read on 2026-11-08.

CapCut's free plan includes free templates and basic editing tools. Pro is $19.99 a month and Ultra is $79 a month, with Teams at $12.49 a month per seat, read on capcut.com/pricing on 2026-11-08. That page says displayed pricing is for reference only and you need to sign in to see final pricing, taxes, and any offers on your account. CapCut doesn't spell out which AI features sit behind Pro, so open auto captions and auto reframe in the app and see what your account gives you before you plan a workflow around them.

How to do it

1. Scrub your footage and write the log. One line per clip. Don't clean it up or make it sound good.
2. Open Claude or ChatGPT in a new chat and paste the prompt below, with your log pasted into the [FOOTAGE LOG] field.
3. Read the three video plans it returns. Each one gives you a working title, a hook for the first three seconds, an edit table with clip file, timestamp, duration, what's said, and what's on screen, plus caption copy written separately for Instagram Reels, TikTok, and YouTube Shorts.
4. Open CapCut and import only the clips the plan names. Nothing else. That alone cuts your edit time in half.
5. Trim each clip to the timestamps in the table. Run auto captions. Type the on-screen text cards at the beats the plan marks, keeping each under eight words.
6. Use auto reframe to export each size you need, then paste the platform-specific caption into each post.

Variables to swap

[BUSINESS TYPE AND NAME], [REGION], [BRAND VOICE], [PRODUCT OR MOMENT], [FOOTAGE LOG], [GOAL], [CTA].

The [GOAL] field does more work than people expect. "Drive tasting room reservations" and "sell through 68 cases of rosé before patio season" produce very different videos from the same footage.

What good output looks like

The edit table should name real timestamps from your log, not invented scenes. If it describes a shot you never filmed, the log was too thin and you should say so and ask it to rebuild from what you actually have.

The first three seconds of each video should say something only your business could say. "Come visit us in wine country" is a failed hook. "This is the last barrel of the 2024, and it's going in the club shipment" is a real one.

The prompt also asks the AI to push back: which of your clips aren't worth using, and the one missing shot that would make these videos noticeably better. Read that part. It's usually the most useful paragraph in the whole output, and it tells you what to film next time.

What to watch out for

The AI is working from your description, not your video. If your log says "Maria pours the Albariño" but the shot is shaky and her face is out of frame, the plan will still call for it. Trust your eyes over the plan when they disagree.

Most people watch social video with the sound off, which is why auto captions and the on-screen text cards matter more than the audio. The prompt asks whether each video works silent. If the answer is no, either add text or pick a different clip.

One more thing: get consent before you post footage of staff or guests, especially in a tasting room where people didn't come expecting to be filmed.

Make it a habit

Save your footage log format as a note on your phone. Every time someone on your team shoots something worth keeping, they add a line. Once a month, paste the whole log in and run the prompt again.

If the first pass feels generic, paste the three hooks you hate back into the chat and tell it to rewrite them using one specific detail from your log. It almost always gets sharper on the second try, because you've shown it what you actually sound like.