PSD, adjustment layers
and local AI cutout.

Studio 3.0 adds practical layer exchange and on-device processing. Read the compatibility report before replacing a Photoshop workflow.

PSD import and export

Import a PSD through Open or Import PSD. The report counts layers, text, groups, masks, smart objects and adjustments. Common RGB raster layers, positions, visibility, opacity and blend modes can be imported separately. Text and smart objects use their cached raster appearance. Group paths become layer names. Supported brightness/contrast, saturation and gamma adjustments become editable Studio layers; unsupported adjustments and advanced masks can look different.

Choose Merged preview when the PSD contains a suitable composite and its saved appearance matters more than editable components. Your original file remains available through Download original PSD during the same session. Imported PSD files are limited to 100 MB and 16 megapixels.

Export writes an 8-bit RGB PSD with separate layers and a composite preview. Simple horizontal Studio text includes editable text metadata; effects are rasterized. Adjustment metadata is included for supported types. Photoshop may render fonts, blending and adjustments differently. LAB, CMYK fidelity, 16/32-bit writing, complex smart objects, advanced text and all Photoshop effects are not fully supported. Keep original files and JSON project backups.

Real adjustment layers

Add brightness/contrast, saturation or gamma from Adjustments. A parameter layer changes the composite of visible layers below it while leaving their source pixels intact. Double-click the adjustment layer or choose Edit adjustment to revise values. Hide it, lower its opacity or move it through the stack to compare results. JSON project save and local checkpoints retain these parameters.

AI background removal

Select a raster image layer, then choose AI cutout. General-object mode uses the Apache-licensed ORMBG model (about 44 MB). Portrait mode uses MODNet (about 26 MB). The ONNX runtime adds about 11 MB. These files download from this site on demand, with no visitor-image upload. A background worker performs the inference locally.

After processing, inspect the preview. Applying creates a new cutout layer and preserves the original as a hidden layer. Undo restores it. Hair, translucent objects, thin structures and complex backgrounds may need manual refinement. Model quality is not guaranteed for every image. Cancel terminates a running worker. The browser and available memory determine processing speed; inference is limited to 16-megapixel source images.

Five language versions

English, Simplified Chinese, Spanish, German and French have separate landing pages and guides. The editor language selector changes interface language without navigating away from the current canvas. User-authored layer names and artwork are not translated.

Privacy and local processing · Software and model notices · Image and project formats