Photoshop’s Generative Fill is no longer limited to the beta workflow shown in this 2023 video. It is now part of current Photoshop releases, uses generative credits, and provides a model picker for Adobe Firefly and supported partner models. The archived tutorial remains useful for the core selection-and-prompt concept; follow the updated steps below for the current interface.
Watch the archived tutorial on YouTube
Open a supported Photoshop version
Update Photoshop through Creative Cloud, open an image, and duplicate the original layer or keep an untouched copy. The earlier requirement to install Photoshop Beta is obsolete for standard Generative Fill, although experimental features may still appear in Beta first.
Make a precise selection
Use the Selection Brush, Lasso, Object Selection, or another selection tool. Include enough surrounding context for natural blending but avoid unrelated areas. Feather or refine difficult hair and edge selections before generating.
Choose a model
The current Contextual Task Bar can offer Adobe Firefly models and supported partner models. Model availability can depend on version, region, account, and plan. Record which model produced an approved result because different models may interpret the same prompt differently.
Add or replace an object
Select the target area, choose Generative Fill, and describe the object, material, perspective, lighting, and position. Generate variations and compare geometry, shadows, reflections, and scale. Photoshop places the result on a Generative Layer, keeping the edit nondestructive.
Remove content
Select the unwanted object plus its shadow or reflection and leave the prompt blank, or use the Remove Tool. Inspect repeated textures and edges. Cleanup may still require masks, cloning, or healing tools.
Expand the canvas
Use the Crop tool to extend the canvas and select Generative Expand. A blank prompt lets the model infer surrounding content; a prompt can guide the scene. Treat invented details as creative material, not evidence of what was outside the original frame.
Use a reference image
Reference Image can guide an object or the whole scene. Upload only media you have permission to use. Match the reference intent—placing a new object or swapping an existing one—and verify that protected logos, characters, or identities were not copied unintentionally.
Credits and commercial use
Generative features consume credits at rates shown in the app. Adobe’s terms can differ by model, plan, and release status, especially for partner or beta models. Check the current usage terms for the exact model before commercial delivery.
Final QA
- Inspect at 100% for seams, extra fingers, warped text, and broken reflections.
- Verify product details and factual elements.
- Keep the original and layered PSD.
- Document prompts, models, references, and permissions.
- Disclose synthetic edits when they could mislead.
Related resources
See our HeyGen guide, vidIQ guide, Filmora link, the AI Tools Arena directory, and the original related video at this link. The creator channel is How To In 5 Minutes.