Getting Started

This guide is written for non-technical users: short steps from upload to final save.

1) Open the app and choose your goal

If you are unsure, use the mode assistant inside the editor. The current modes fit well into 3 groups:

Improve an existing photo

Single image edit: precise local changes with prompt and optional markup.
Anime conversion: turn a photo into anime/cartoon style while preserving identity and colors.
Background edit (fast): change or restyle only the background.
Cutout: remove the background and return the subject as a transparent PNG.
Clean up scene: remove selected distractions while preserving the main subject.
Relight photo: change only lighting, shadows, and mood without moving the scene.
Enhance photo quality: improve rendering and sharpness, including recovery of slightly blurry or motion-blurred photos.
Upscale image: genuinely enlarge the final image with a factor from 1x to 4x.
Product photo mode: create packshots and commercial product images from a product photo.
Real estate listing photo: correct light, perspective, and clutter while preserving the real property.
Restore photo: repair old or damaged photos.

Combine multiple images

Compose multiple images: merge 2 to 6 images into one scene.
Virtual dressing room: apply a garment to a subject using a subject photo plus a garment photo.
Seamless Blend (subjects): place extracted subjects manually into one scene.

Create from scratch

Generate image from text: no source image required.
Sketch to image: start from a drawing.

2) Upload the right input

Edit/Anime/Background/Relight/Enhance/Restore/Upscale: upload 1 image.
Cutout: upload 1 image. It works best with a clearly readable subject separated from the background.
Clean up scene: upload 1 image, select the distractions to remove, and add any custom items under Other.
Product: upload 1 product photo. You can enable an optional second image as a background/environment reference.
Real estate listing photo: upload 1 sharp room photo and use conservative settings.
Compose: upload 2 to 6 images. After 6, you must remove one before adding another.
Virtual dressing room: upload 1 subject photo + 1 garment photo. You can enable an optional third accessory image.
Seamless Blend: main image optional + up to 5 subject photos; to generate, at least one subject must be placed on the canvas.
Generate: no upload required.
Sketch: create or upload the sketch you want to render from.

Tip: use sharp inputs, avoid heavy compression, and keep lighting compatible when combining multiple images.

3) Know when the prompt is actually needed

The prompt is not mandatory in every mode:

It is usually required or strongly recommended in Single image edit, Background edit, Generate image from text, and Sketch to image.
You can start even with no prompt in Product photo mode, Enhance photo quality, Upscale image, Restore photo, Compose multiple images, and Seamless Blend: the app adds internal default guidance.
In Relight photo, there is no free-form prompt: you use templates + the builder in Advanced controls.
In Anime conversion, there is no free-form prompt: choose a style and run the conversion.
In Virtual dressing room, there is no free-form prompt: if you need constraints, use the structured builder in advanced controls.
In Cutout, you do not need a prompt at all: the mode uses a dedicated workflow that removes the background automatically.
In Clean up scene, you do not need a free-form prompt: choose distractions and use Other for specific items.

When you do write a prompt, use this formula:

what to change;
what must stay unchanged;
desired style.

Quick examples:

Edit: “Remove the road sign behind the subject. Keep face and pose unchanged. Realistic look.”
Compose: “Use the living room from the first photo as the main scene and place the dog from the second photo near the sofa, with realistic scale and matching light.”

4) Choose the right model

You do not need to study every model before you start. Use this practical rule:

Model	Choose it when
Standard	you want the lowest cost and a solid first result
Standard+	you are combining several images and need stronger consistency
PRO	the content or workflow needs a more capable model
PRO+	accuracy, consistency, and instruction-following matter most
MK	you are in Product mode and want the most accurate commercial result

The app only shows models compatible with the selected mode and options. When in doubt, start with Standard; move up only if the first output does not preserve identity, composition, or instructions well enough.

Cutout has a separate choice:

Dynamic: the general option for most images;
Matting: useful for fur, background hair, and complex edges;
Portrait: best for close portraits where hair is prominent.

5) Run the first result

Click the main action button in your selected mode.

Useful notes:

In Relight photo, start from a lighting template close to your target and raise background adaptation only when needed.
In Relight photo, open Advanced controls to choose the template and tune the lighting plan.
In Anime conversion, pick a style and run the conversion; the subject and colors stay faithful to the source.
In Cutout, output is always a transparent PNG and Variations are not available.
In Clean up scene, the model identifies the main subject and removes only the selected distractions.
In Product photo mode, choose the preset first and add a background reference only if you really need your own set or environment.
In Enhance and Restore, start from the softer preset first; if the photo is blurry or motion-blurred, move to Balanced or Aggressive only if Soft is not enough.
In Upscale image, start from 2x if you are unsure: higher factors increase both cost and reinterpretation risk.
In Virtual dressing room, there is no free-form prompt: use the builder in advanced controls only when you want to guide fit, styling, accessories, or preservation rules.
In Seamless Blend, the manual composite is harmonized by AI only when you launch the final action.

6) Create variations and pick the best

Use Variations (2-4) to get alternatives without rewriting your prompt.

7) Improve with “Use as Input”

When you find a good variation:

click Use as Input;
run a second pass;
fix small remaining issues.

In practice:

after Product photo mode, refine any small defects or tricky edges in Single image edit;
after Upscale image, if you see small local artifacts, refine them in Single image edit;
after Cutout, you can reuse the transparent PNG as the base for a new layout or place it into another scene;
after Compose or Seamless Blend, final cleanup often happens in Single image edit;
after Virtual dressing room, use Edit only for small local artifacts.

8) Save your final image

Download/Save: save on your device.
Save to Drive: save in Google Drive.
Share: available on supported devices.

Note: variations are not auto-saved, so save them manually.

9) Working on many images

Use Batch mode only in Enhance and Restore:

enable Batch mode;
upload up to 20 images;
run and review outputs in Gallery.

Batch requires Google Drive connected and Drive auto-save enabled.

10) Common issues

It changed too much

add clear constraints in your prompt;
in Virtual dressing room, set preserve rules in the builder (face, pose, background, proportions).

The merge looks unnatural

in Compose or Seamless Blend, ask for realistic scale, coherent lighting, and believable shadows;
choose inputs with compatible perspective.

You see artifacts

create variations;
pick the cleanest one;
run a second pass with Use as Input.

Fine cutout edges are not perfect

use a sharper source image with stronger subject/background separation;
if needed, clean up the exported result in Single image edit.

The outfit does not fit well

in Virtual dressing room, use the builder to guide fit, fabric, and silhouette;
use a clear garment reference photo.