Adobe Stock

Adobe Stock offers a Find Similar feature that uses an AI image-similarity engine to surface stock assets that visually match a selected image, allowing users to search by content, color, or composition instead of — or in addition to — text keywords.

The Find Similar system accepts an image (either from Adobe Stock search results or uploaded by the user) as input and produces a ranked list of visually similar stock images and videos as output. The underlying technology uses machine learning models that convert images into dense numerical vectors (embeddings) and then compute similarity scores across the collection. Users can refine results by sub-filters: Content (similar subject matter), Color (matching palette), or Composition (similar layout and depth). Adobe's engineering blog described the system as built in partnership with Adobe Research and the Adobe Sensei team.

Adobe Stock offers AI-powered editing tools, branded as Generative Edits and Generative Variations, that let all stock customers modify existing licensed images before downloading, using features such as background removal, background replacement, image expansion, and style or composition changes. These tools were first made available to business customers in 2024 and became generally available to all stock customers in November 2024.

These tools are powered by Adobe Firefly and operate on existing Adobe Stock images as a reference. Inputs include a selected stock image plus optional text prompts describing the desired change; outputs are modified or newly generated images reflecting those changes. Specific capabilities include Replace Background (swaps the background via text prompt), Expand Image (extends the image to a new aspect ratio), Style (applies the aesthetic of one stock image to a new generation), and Composition (matches the layout and depth of a reference image). Generating an edited image consumes a generative credit; downloading the resulting image may also consume a stock credit depending on plan type. Contributors whose stock images serve as references receive compensation when the generated output is downloaded.

Adobe uses licensed images, vectors, illustrations, and videos contributed to Adobe Stock as training data for its Adobe Firefly family of generative AI models, and pays eligible contributors an annual bonus tied to the volume and licensing performance of their assets used in training.

Adobe confirmed that Firefly's first commercially released model was trained on Adobe Stock licensed content, openly licensed content, and public domain content where copyright has expired. Contributor agreement terms permit this use. Adobe pays contributors a discretionary annual Firefly Contributor Bonus: the 2023 bonus covered data used from June 2022–June 2023; the 2024 bonus added videos for the first time; and the 2025 bonus covered data from June 2024–June 2025. Bonus amounts vary per contributor and are weighted by the number of licensed assets. Adobe confirmed in its official Firefly FAQ that it does not allow contributors to opt out of having their content used for training.