If you sell dozens or hundreds of products, making a short video for each by hand is out of the question, and a template filled with different photos produces fifty reels that feel like one. Followers notice. This post looks at how one open-source generator, motion-promo-reels, deals with that, and what it means for a small shop thinking about a catalogue-to-reel workflow. I have read its README and skill file; I have not run it on a store.
What it does for a catalogue
For shops with many products the project has a catalogue mode: vertical 1080×1920 video, 15.5 seconds per product, one reel per product, with one hand-written gag per product category. It can read product data straight from a WooCommerce store; the skill file also mentions Shopify as a data source and notes that WooCommerce prices come in cents. Prices, colours, sizes and features are taken from the website or store API, not typed or invented.
What it varies on purpose
The README describes a seeded planner that gives every video its own construction. For the catalogue mode, the options it lists are:
- 6 scene orders
- 3 product-scene layouts
- 3 range-scene layouts
- 3 facts layouts
- 3 price layouts
- 6 transitions
- 3 colour themes (light, dark, accent)
Per the skill file, the plan also covers text entrance, hook variant, music style, tempo, key, sound-effects pack and seed. Audio is synthesised, with 10 music styles and 5 sound-effect packs. The planner tries to avoid repeating recently published constructions and reports the minimum distance between neighbours in a VARIACION.md file; the target for same-construction pairs is zero.
Do the sums on what that means: with just the layout, transition and theme choices above, there are 6 × 3 × 3 × 3 × 3 × 6 × 3 = 8,748 possible visual combinations before music is counted. Fifty reels from 8,748 options should rarely collide, and the planner checks anyway.
What it keeps fixed
Variation is not the same as chaos. The skill file says the signature element, closing card and key audio sync points never vary, and that brand rules override preferences. The 15.5-second structure stays: a hook, the product, a range scene, facts, price and an outro, with the three middle scenes allowed to reorder. That fixed spine is what makes fifty reels recognisably yours.
The checks a shop owner should copy
You do not need this tool to borrow its checklist. Before posting any batch:
- Confirm today's price. The skill file asks for the date the data was read, because prices go stale. Check sale expiry and units too.
- Use “from” wording for price ranges. One reel should not promise a price only the cheapest variant has.
- Look at stills, not just the video. The tool makes contact sheets so you can scan for overlapping text, washed-out letters and empty blocks.
- Keep key text inside the safe area. The skill keeps nothing important in roughly the top 250 px or the bottom 320 px of the frame; these are the project's own working margins, not an official Instagram specification.
- Label AI-made people. If your product photos show AI-generated or AI-modified people, the skill adds a visible label and says to disclose the same when publishing.
- Skip what you cannot confirm. Unconfirmed units or features go into a review file instead of the video.
Honest limits
- It is a developer tool. It needs Node 18+, Chrome or Edge, ffmpeg, and Python 3 with numpy, scipy and Pillow. There is no point-and-click app. A shop owner would likely need a developer or an AI coding assistant to set it up.
- It is small and young. The repo showed 6 stars and 15 commits when I looked. Test on five products before trusting it with five hundred.
- Variation does not equal quality. Different layouts do not make a weak product photo sell. Photos still come first.
- Gags need a person. The skill wants each gag to be a real, checkable property of the product. It also says never to joke about bodies and to use a calm tone for sensitive categories. Someone on your side should read every line.
Should you try it?
If you have at least 30 products, a developer friend, and a store API to read from, a pilot of five reels is a cheap experiment: generate them, read the variation report, check the stills, post them, and see whether people watch past the first three seconds. If you have a dozen products, make them by hand.
