Most stock portfolios are smaller than they should be, and the reason is rarely a shortage of content. It is a folder of finished work that never got tagged. This page maps the full contributor workflow from selection to sale, what breaks at each stage, and which parts are worth automating once volume grows.
The workflow in order
Select and cull → generate metadata for the batch → review and correct → embed IPTC → export per-agency formats → deliver over FTP → verify ingest → next batch.
The order matters more than it looks. Metadata goes in before delivery, because fixing it afterwards means editing file by file through a web interface — which is exactly the work batching was supposed to remove. Adobe Stock in particular has no CSV import route at all, so anything not embedded before transfer has to be typed into the portal.
Stage 1: Selection
This stage stays manual, and it is where approval rates are won or lost. Submitting an entire shoot means submitting your weakest frames, which invites similar-content rejections that can affect the whole batch. Choose the strongest frame per setup. Our guide to raising approval rates covers what reviewers actually reject and why fewer files often earn more.
Stage 2: Metadata — where workflows die
Manual keywording costs five to twenty minutes per image. At that rate a 500-file batch is several weeks of work, which is why most contributors simply do not do it. Worse, quality degrades measurably: by roughly the twentieth image most people are recycling safe terms and writing thinner descriptions, so the tail of every batch gets weaker metadata than the front.
Start with the fundamentals: what IPTC, EXIF and XMP actually are and which fields agencies read, then how to write metadata and how to choose keywords that convert. Underneath all of it sits how buyers actually search — for a use rather than a subject, which is the single most common gap in keyword lists.
Stage 3: Review — the part automation cannot replace
Every vision-based tool shares the same blind spots: specific place names, editorial dates and events, precise species, brand identification, and the concept you intended in an abstract shot. The workflow that holds is AI-first, human-reviewed — let software do the exhaustive part, then spend seconds correcting what only you know. Our guides to AI metadata generation and bulk metadata generation cover where it holds up and where it needs supervision.
Stage 4: Per-agency formatting
One keyword list does not fit every agency, and this is where multi-agency contributors lose files. Adobe accepts 49 keywords and reads only embedded IPTC. Shutterstock accepts 50 plus categories from a fixed list, applied by CSV after upload. Getty and iStock map to a controlled vocabulary and refuse externally generated AI. Freepik requires a mandatory AI flag. Alamy has no keyword cap but banned AI submissions from 1 September 2026.
Our Shutterstock CSV template guide covers one platform's exact column structure, and category mapping covers the field contributors most often get wrong.
Stage 5: Delivery
Above a handful of files, FTP replaces the web uploader: it resumes interrupted transfers, handles large batches without a browser session, and can be automated. Three settings cause most silent failures — transfer in binary rather than ASCII, upload to the root rather than subfolders, and keep concurrency modest. Our Adobe Stock FTP guide and bulk upload guide cover the setup and the errors that lose files without an error message.
Video and vectors need extra stages
Footage carries fields stills do not — camera movement, shot type, frame rate, duration — and editors filter on them before browsing. See footage metadata and tagging a shoot as a series. Vectors add a structural pre-flight: outlined text, expanded strokes, no embedded rasters. See vector requirements.
What to automate and when
Under a hundred files a month to one or two agencies, automation is overhead — agency-native keyword suggestions will cover you. The calculation changes when per-file time multiplied by your volume exceeds what software costs, which for most contributors is somewhere between two hundred and five hundred files a month, or the point at which you add a third agency.
Our comparison of keywording tools covers the options honestly, including where each one is the wrong choice. Rastock AI generates per-agency metadata, embeds IPTC and delivers over FTP to 10+ destinations in one pass, with plans by upload volume on the pricing page.
Once the workflow runs, the next question is where to send the output. Our guide to where to sell stock photos compares every major agency on royalties, review standards and AI policy.
Related reading
Bulk Metadata Generation: How to Tag Thousands of Files Without Losing Quality
AI Metadata Generator for Stock Photos: Why Manual Tagging Is Dead
Bulk Uploading to Vecteezy: FTP Delivery and Batch Metadata
Frequently asked questions
What is the correct order for a stock photography workflow?
Select and cull first, then generate metadata for the batch, review and correct it, embed IPTC into the files, export per-agency formats where needed, deliver over FTP, and verify ingest before starting the next batch. Metadata always goes in before delivery, not after.
Where do most stock workflows break down?
At metadata. Writing a title, description and up to 49 keywords takes five to twenty minutes per image manually, so a few hundred finished files become weeks of work. That is why most portfolios are far smaller than the amount of content their owner has actually produced.
Should I embed metadata or use CSV?
Both. Embedded IPTC travels inside the file to every agency and is the only bulk route Adobe Stock offers. CSV handles platform-specific fields IPTC cannot carry, such as Shutterstock category assignments and Freepik's AI flag.
How many files should I process in one batch?
Enough that automation pays off, few enough to verify before committing more. A few hundred is usually manageable — you can confirm ingest and metadata landed correctly, so a systematic error costs you three hundred files rather than eight thousand.
Which parts of the workflow should stay manual?
Selection and the review pass. Culling requires judgement about what is worth submitting, and reviewing generated metadata catches the things software cannot know — place names, editorial dates, species, brands and the concept behind abstract work.
Is FTP better than a web uploader?
Above a handful of files, yes. FTP resumes interrupted transfers, handles large batches without a browser session and can be automated. Web uploaders are fine occasionally and become the bottleneck at volume.