How to Use Fooocus

Fooocus is a Stable Diffusion interface designed for people who want strong image results without tuning every sampler, scheduler, and model parameter manually. This guide is for the search intent behind "how to use Fooocus": what to type, which controls to touch first, and how to move from a rough prompt to a finished image.

If you only want to experiment in the browser, start from the Fooocus Playground. If you want a broader multi-model creation workspace with video, upscaling, and background removal, use the AI image and video generator.

1. Start with a specific prompt

Write the prompt as a production brief, not as a list of random keywords. A useful Fooocus prompt usually includes the subject, setting, composition, lighting, style, and quality target.

Example structure:

portrait of a ceramic artist in a bright studio, hands covered in clay, soft window light, shallow depth of field, editorial photography, natural skin texture

Keep the first test prompt short enough that you can tell which change improved the result. Add detail after you see the first batch.

2. Choose style controls before advanced settings

Fooocus is useful because style presets can do a lot of the heavy lifting. Before changing advanced values, test the closest style direction: photographic, cinematic, anime, illustration, product render, or concept art.

Use one prompt and change only the style preset for the first comparison. This keeps the test clean and helps you learn what each preset actually changes.

3. Generate variations and read the failures

Run a small batch, then look for repeated issues:

• If the subject is wrong, rewrite the first half of the prompt.
• If the style is wrong, change the style preset before adding more keywords.
• If the composition is wrong, add camera angle, crop, or framing language.
• If details are messy, simplify the prompt and generate again.

Do not keep adding more words to a weak prompt. Fooocus usually responds better when the main idea is clearer.

4. Use image prompts when you need visual direction

When text is not enough, add an image prompt or reference image. This is useful for pose, composition, character direction, product shape, or general color mood.

Use references for structure and prompts for intent. For example, the image can define the pose while the text defines the clothing, lighting, and final style.

5. Fix local problems with inpainting

Use inpainting when most of the image works but one area needs correction. Mask only the part that should change, then write a prompt for that area.

Good inpainting prompts are specific:

• "replace the left hand with a natural relaxed hand"
• "clean product label with readable minimal typography"
• "remove the extra chair in the background"

Avoid asking inpainting to redesign the entire image. If the main composition is wrong, return to the generation step.

6. Upscale only after the image is selected

Upscaling is a finishing step. Pick the best candidate first, make any inpainting fixes, then upscale for export. This keeps the workflow faster and avoids spending time enhancing images you will not use.

For site graphics or product visuals, check the final image at the size where it will actually appear. A picture can look good in preview but fail when cropped into a banner, card, or social format.

7. Save the prompt that worked

When you get a good result, save the prompt, style choice, seed if available, and any reference notes. This makes the image repeatable and helps you build a small prompt library for future Fooocus sessions.

Quick Fooocus workflow

1. Write one clear prompt.
2. Test style presets before advanced settings.
3. Generate a small batch of variations.
4. Rewrite the prompt based on repeated failures.
5. Use image prompts for pose or composition.
6. Use inpainting for local fixes.
7. Upscale the selected final image.

Common mistakes

• Starting with a vague prompt like "beautiful art" and hoping the style preset solves everything.
• Changing prompt, style, and settings at the same time, which makes results hard to diagnose.
• Using inpainting for a full redesign instead of a local correction.
• Upscaling too early.
• Treating Fooocus, the Playground, and the AI generator workspace as the same search intent.