Run an AI vector-art stock asset pipeline but keep four human checkpoints: keywords, upload go-ahead, image review and pricing
A solo seller produced 39 vector asset packs for Adobe Stock, Vecteezy and Dreamstime with an AI prompt engine and Python scripts, and explains which steps they automated and which stayed manual after costly mistakes.
Evidence: The author reports this. We have not checked it beyond reading the source.
The business problem
AI image generation makes it cheap to produce stock asset packs in bulk, but bad keywords, duplicate uploads and flawed images get rejected by marketplaces or caught by buyers.
What was tried
1. Generate vector art with a prompt engine (the AI model is not named) and package each product with a script that renders a 1280x720 cover and a 600x600 thumbnail and zips the files. 2. Generate the metadata CSV for each marketplace from one command (Adobe Stock, Vecteezy and Dreamstime each want different columns). 3. Keep an upload ledger script that shows what is pending, and treat it as a manual go/no-go gate, because Adobe does not warn about duplicate uploads. 4. Write titles and keywords by hand or from a prepared file, because the script's defaults were poor and bad keywords lead to failed reviews or no sales. 5. Review every image by eye for broken hands, paths and artefacts, about two hours per 39-product batch. 6. Decide product groupings, price and target buyer from research notes, and keep the research folder permanently. 7. Run everything on a small machine (1 vCPU, 2 GB RAM, no GPU) by processing one asset at a time and rendering previews at low resolution.
What was reported (mixed)
Self-reported: 39 products (856 MB) built with about 19,000 lines of pipeline code. Mistakes that shaped the checkpoints: one 40-vector batch was uploaded twice and flagged as duplicates (about six hours of upload work lost and two days of cleanup); skipping image review twice let buyers find the flaws first; deleting the research folder lost material for two products. No sales, revenue or rejection rates are reported.
Limitations
The author sells a vector prompt pack and a pipeline starter kit containing these scripts, so the post is partly promotion. The AI image model is not named, and no sales or acceptance figures are given, so it is unknown whether the products earn money. Single seller, self-reported.
What you need
An AI image or vector generation setup, contributor accounts on the stock marketplaces, and basic Python.
Sources
- DEV Community (Muhamad Revanza Rafif Hamdani) ↗ Vendor case study, published October 21, 2026
Source published: October 21, 2026. Last reviewed here: October 21, 2026. Spot a mistake? Tell us.
Tools in this workflow
- Adobe Stock: MarketplaceTry Adobe Stock
- Vecteezy: MarketplaceTry Vecteezy
- Dreamstime: MarketplaceTry Dreamstime
Get new workflows like this by email
A short selection of the most useful new workflows, including the ones that did not work. Free, and you can unsubscribe at any time.
Related workflows
Shipping 43 Chrome extensions with Claude Code: what the numbers showed and why 22 came down
A solo developer used Claude Code to build 43 Chrome extensions from AI-suggested ideas, then published every number: 22 taken down, about $107 plus 3,480 yen in external revenue over nine months.
Connect an AI agent to a merchant-of-record so it can create products and checkout links without owning tax
A vendor tutorial for letting Claude or another agent create a product and a checkout link through a merchant-of-record, so VAT and sales tax are not the founder's filing job.
Seed an AI agent with $50 of Bitcoin to build and sell a pay-per-use product, and what failed
A company blog post tests whether a master AI agent and sub-agents can build and launch a small paid web product alone; it reached zero revenue and the agent overstated progress.
Let an AI agent run a SaaS alone, then audit it: what an autonomous build got wrong
An experiment gave an AI agent full control of a small SaaS build, then audited it and found human-style mistakes: too many products, too many offers and overconfident content.