At CES on 8 January 2024, Getty Images launched Generative AI by iStock for small businesses, designers and marketers. Its positioning and buyer context differed from Getty’s earlier enterprise-targeted generative offering.
The January announcement
Getty described the iStock tool as built on NVIDIA Picasso and trained on content from Getty’s creative libraries. The supplier said the system was engineered to guard against some known products, people and locations. These are vendor design claims, not independent evidence that every output avoids protected objects or identifiable people.
A practical small-business brief
Suppose a bakery needs a decorative campaign scene. It should define palette, season, aspect ratio, empty title area and desired atmosphere while avoiding claims that a synthetic cake exactly matches the actual item sold. An owner should use real product images for the menu or ordering page and synthetic pictures only where they are unambiguously conceptual.
Separate a generator from its editing capabilities
Launch materials discussed generated artwork and visual design rather than universal accuracy guarantees. Where an interface offers further editing, distinguish re-generating a region from compositing actual supplier assets. Record what is new generated content and what originated from an authorized photograph. A mask-based edit should not silently change the brand’s real product form.
Commercial-safety language has limits
Getty’s launch described legal protection and rights as part of the offering. Contract terms, country, intended medium, brand identifiers and other factors still matter. Never infer that a priced license covers a celebrity, third-party logo or intentionally copied living artist’s recognizable identity. Record the actual license and any indemnity clauses before classifying a customer campaign as cleared.
Visual checks that a small team can perform
Review generated props and hands at full size, verify any printed numbers and packaging, then inspect the final design on mobile and desktop. Have a second person check whether the visual might be mistaken for genuine documentary evidence. Keep prompt and output versions in a folder with a human-signed approval note and a link to the license record.
Getty versus iStock is a procurement distinction
The vendor’s September 2023 Getty Images tool and January 2024 iStock tool should have separate catalog entries and dated descriptions. A team should select whichever actual license and workflow match its work rather than assuming identical entitlements or identical technical models across the two branded services. A specialized product photo workflow may still be preferable for SKU fidelity.
Worked scenario: a hypothetical production brief
Example: A small bicycle retailer wants a spring-sale banner. The actual bicycle SKU is photographed, retaining real component count and frame geometry. A synthetic background may add an atmospheric path or sky, but should not make a nonexistent bicycle specification appear authentic. The retailer checks masking boundaries around the product, verifies the license for the generated visual and leaves pricing in editable text. A final review compares the advertised item with the real inventory image. The process demonstrates that synthetic backgrounds and verified merchandise photography can coexist without rewriting the product.
Additional editorial and quality safeguard
A small business rarely needs the highest volume of generated variants. A better starting project is one actual campaign, two visual routes and one human approval. This limits the risk of publishing inconsistent depictions across social accounts and storefronts. Track the final license receipt, and archive the exact asset that was approved rather than substituting a later regeneration.