Short answer
Product photography studios hold large, consistently lit image archives with shot lists and file names tied to client SKUs. That makes them attractive for object recognition, but the studio often does not own the rights: clients, brands or photographers may. Licensing means tracing each image's rights chain. The images show packshots, not workplaces or handling. This is not legal advice.
Why studios are a distinct supplier type
A commercial product studio produces exactly what a recognition model wants and a warehouse camera does not: a known object, isolated, repeatably lit, shot from agreed angles, often with a clean cut-out and a record of which SKU it is. Over years, a busy studio accumulates a large archive across many brands.
Compared with retailers and manufacturers, studios sit one step away from the product. They did the work, but a client commissioned and paid for it. The agreement between them decides who can license the image for a new purpose such as model training, and that agreement was almost certainly written before anyone considered training. Buyers who treat the studio's hard drive as the asset can end up with files nobody can lawfully license.
For how brand and photographer rights stack up, read product image rights. The cluster overview is the product and object data hub.
What a studio archive typically contains
Archive composition varies by studio, so each item below is a question for the studio, not a promise:
- Packshots. Front, side and back views on white or a neutral ground, often with clipping paths or masks.
- Multi-angle sets. Turntable or rotated views of the same item. See the multi-view product photography spec.
- Lifestyle images. Products in staged scenes, with models and props. These carry people and extra rights.
- Detail shots. Labels, closures, textures, ingredient panels.
- Working files. Raw captures, color charts, lighting notes, retouching layers.
- Job metadata. Client, SKU, shoot date, photographer, usage terms from the original order.
The working files and job metadata are often more valuable to a perception team than the delivered JPEGs, because they show how the image was made. Color accuracy and lighting records matter when color is a label; see color calibration in product images.
The rights chain, step by step
Rights questions come before image quality. Treat these as diligence prompts, and have counsel review the answers. This is not legal advice.
- Who made the image? Staff photographer, freelancer, assistant, or an outside retoucher? Employment and contractor terms differ.
- Who commissioned it? The brand, an agency, a retailer or the studio's own speculative shoot.
- What did the original order allow? Many studio agreements grant the client a defined use, such as web and print, and keep other rights with the photographer or studio. Some assign everything to the client.
- What is in the frame? Trademarks, packaging design, artwork, other brands' goods, hands or faces of models.
- Is third-party licensed content inside? Stock backgrounds, fonts, licensed prop imagery, or CC-licensed material.
- Can the studio license for model training? If the studio cannot, can the brand? If neither, a separate agreement may be needed.
The U.S. Copyright Office's pre-publication report on generative AI training describes the points in AI development where copying may implicate rights and examines licensing options (U.S. Copyright Office, 2025). It is background reading for the questions above; it does not resolve any particular archive's position.
If an archive includes Creative Commons material, read the actual license. NonCommercial is judged by the intended use, not the user, and ND terms restrict public sharing of adaptations; licenses with NC or ND terms are considered non-free (Creative Commons). Whether a given term fits a commercial training license is a legal question for the license text and counsel.
Illustrative archive evaluation sheet
Illustrative only; the values are synthetic and the columns are one way to organize a review, not a standard.
| Field | Illustrative value | What it settles |
|---|---|---|
job_id | job_5530 | Ties files to an original order |
client_type | brand | Who may hold commissioning rights |
usage_terms_on_file | web_print_2y | Whether training was ever contemplated |
photographer_status | freelance | Whether an assignment or license exists |
contains_people | no | Separates packshots from lifestyle shots |
third_party_marks | brand_logo_on_pack | Trademark questions |
licensable_for_training | unknown | The field you need to resolve before purchase |
A first pass often shows that only part of an archive is licensable. Budget time for the rest to be cleared, excluded or replaced.
Existing archive versus a new studio shoot
Where the archive is not licensable, or does not match the SKUs or angles you need, a new shoot under work-for-hire or assignment terms is the alternative. A studio is well placed to run it: you set the capture spec, calibration and QA, and the agreement is written for training from the start. The cost is time, and the images then cover only what you commissioned. The decision is laid out in existing assets, new collection or specialist partners.
Ask SourceX whether a commissioned shoot is feasible for your objects; scope and cost are case by case.
What studio archives cannot support
Packshots are idealized. They carry no clutter, no occlusion, no wear, no shelf context and rarely any scale reference. A model trained only on them tends to see a world where every object is centered, well lit and alone. Studio images do not show how an object feels, weighs or deforms, and they contain no robot sensor data. Training a humanoid to grasp or place products needs robot or teleoperation data, not product photography. See object recognition versus manipulation data and packshot versus in-situ images.
Other limits:
- Brand skew. A studio's clients are not a sample of your store's products.
- Retouching. Edits can change shape, color and labels, so the pixels may not match the physical item.
- Age. Old products shot in old packaging may no longer exist.
- Evaluation. Hold out brands or whole shoots, not random images from the same session.
Two related pages help with the clearance work: trademarks in training images for logos and packaging art, and retailer versus manufacturer object data for comparing a studio with the other image-holding supplier types.
Questions to ask suppliers
- Which clients, product categories and years does the archive cover?
- Who took the images, as employees or contractors, and are assignments or licenses on file?
- What do the original orders say about use beyond the client's initial purpose?
- Can the studio license images for model training, or does that require client consent?
- Are raw files, color-chart frames and lighting notes retained and licensable?
- What retouching was applied, and are unretouched versions available?
- Do any images include people, third-party artwork or licensed stock elements?
- Are SKU or barcode identifiers recorded per image?
- Would the studio take a new commission under work-for-hire terms?
Discuss your requirements
If you want to know whether a studio archive could meet your object-data needs, describe the product categories, image characteristics and permitted uses to SourceX. SourceX looks for US businesses that hold suitable material, checks the data and the supplier's licensing permissions, and coordinates the license and delivery. A request does not guarantee a match; SourceX follows up about potential sources. See the humanoid training data guide for the other routes, and the training data due diligence checklist for a general checklist.
Explore sourcing and licensing options with SourceX
Sources
- About CC Licenses — NonCommercial and NoDerivatives terms, relevant to CC material inside archives.
- Copyright and Artificial Intelligence, Part 3: Generative AI Training (pre-publication version) — where copying in AI development may implicate rights; licensing options.