Frequently Asked Questions

Zappa is a prompt engine for AI-generated editorial fashion photography. You upload a reference image (or describe a scene), and Zappa generates a structured prompt that you can use in tools such as Higgsfield. The engine translates pose, clothing, lighting, environment, and composition into precise, repeatable instructions.

Not required. You can also fully describe a scene in the text field. However, a reference image gives Zappa much more to work with: pose, composition, lighting, and material behavior are directly translated from the image. The reference image is used solely as visual inspiration—an image is never copied.

Face and hair are not controlled through the prompt, but via a separate reference image (identity lock) in your generation tool, such as Higgsfield. Zappa protects your identity by deliberately excluding these details from the prompt. The prompt does include a fixed instruction stating that hair, face, and identity must remain exactly as in the reference image.

Yes, as a practical styling instruction. For example: hair combed back, hair out of the face, or tightly behind the ears. The goal is to control the image, not to change your identity. Zappa processes this as physical styling, not as an identity modification.

This is the free input field where you can specify adjustments or preferences. You can use it for two purposes:

• With a reference image: to indicate changes, such as a different pose, environment, or specific styling.
• Without a reference image: to describe an entire scene, including pose, clothing, atmosphere, and setting.

You may write this in Dutch — Zappa will translate it automatically.

The toggles allow you to control which elements are included in the prompt:

• Clothing — If off, Zappa describes the clothing from the reference image. If on ("Add your own clothing"), Zappa skips the clothing so you can provide your own as an attachment.
• Jewelry — Same principle: off = jewelry from the image is described, on = jewelry is omitted.
• Camera & Optics — On/off: whether camera perspective and lens effects are included.
• Environment — On/off: whether the background and surroundings are described.
• Lighting — On/off: whether light sources and shadow behavior are included.

The clothing and jewelry toggles are useful if you want to use your own styling or accessories instead of those in the reference image.

Some AI models automatically add jewelry when hands, ears, or necks are prominently visible. This is due to internal defaults of the model, not the prompt. Zappa offers a jewelry toggle that lets you specify that jewelry should not be taken from the reference image. However, the AI model may still add them — unfortunately, this cannot be completely prevented.

The greatest impact comes from:

• A good reference image with a clear pose and composition
• Specific adjustments in the text field (e.g., stronger lighting, more contrast, different background)
• The correct toggle settings for clothing and jewelry

The more concrete your input, the better the result. Avoid vague terms like "beautiful" or "professional"—instead, describe exactly what you see.

A regular prompt often vaguely describes what you want to see: "a beautiful woman in a studio." Zappa works differently—it describes exactly what should be visible: precise body position, material behavior of clothing, light direction, shadow behavior, and points of contact. Every sentence is verifiable and reproducible. This consistently delivers better and more realistic results.

Zappa deliberately avoids: • Descriptions of faces and hair (identity is managed separately) • Cinematic or narrative language ("she walks elegantly") • Mood and emotion words ("mysterious", "dreamy") • Camera brands or technical metadata The focus remains on what is visible and reproducible: pose, material, lighting, composition, and environment.

1. Generate a prompt in Zappa based on your reference image or description.
2. Copy the prompt (click Copy).
3. Open Higgsfield and paste the prompt into the text field.
4. Add your own face/identity as a reference image in Higgsfield.
5. Optional: Add your own clothing or jewelry as an extra attachment (if you have enabled those toggles in Zappa).

Zappa provides the full description of pose, lighting, environment, and composition. Higgsfield supplies the face and generates the final image.

Upload a reference image and click Generate. That’s sufficient—Zappa extracts everything needed from the image. Want to make adjustments? Use the text field for specific requests such as a different pose, environment, or lighting direction.

You can enter your description in any language—Dutch, English, or any other. Zappa automatically translates everything into the correct format for the prompt. The generated prompt is in English by default (for optimal AI results), but it can also be executed in Dutch.

AI models are naturally cautious and often create "safe" images. Tips to improve this:

• Use the text field for targeted adjustments: "harder light," "more contrast," "more dramatic shadows"
• Try a different reference image with more visual tension
• Small changes in pose or lighting direction can make a big difference

The prompt already includes as many guidelines as possible to achieve a strong result, but the AI model sometimes needs an extra push.

No. Your uploaded images are temporarily stored for processing and automatically deleted immediately after. When you upload a new image, the previous one is also removed. No images are kept on our servers. The processed results (full body and knee crop) are only displayed in your browser — download them directly if you want to keep them.

Correct Prompt is a tool that lets you adjust an existing prompt via AI. Paste your prompt, describe what you want to change, and the engine rewrites the prompt while maintaining the Zappa structure.

Use the Engine to generate a new prompt from a reference image. Use Correct Prompt when you already have a prompt but want to make adjustments — for example, a different background, more dynamic pose, or different lighting.

Yes. The AI rewrites the prompt according to the same sections and rules as the Zappa Engine: framing, pose geometry, wardrobe, hair lock, lighting, environment. Only the requested change is applied.

Yes. You can use the output of a correction as new input and describe further adjustments. Each iteration preserves the structure.

The module processes a Higgsfield image into a studio-ready result. It separates the model from the background using AI masking (Replicate rembg), applies a background tint, adds texture and sharpness, and generates automatic crops. This ensures consistent image quality, making it ideal for your webshop.

No. Uploaded images are deleted from our servers immediately after processing. We do not store any images. The processed results are saved only locally in your browser (up to 16 items) and are not accessible elsewhere.

Reference images follow a strict order:

Image 1 = Face (Identity Lock) — facial structure, skin tone, proportions transferred exactly.
Image 2+ = Garment (Product Only) — clothing and shoes only, no influence on pose/lighting/identity.
Last image = Pose (Only Posture) — only posture and expression, nothing else.

Each reference has a single isolated function. No blending between roles.

The default tint (#f3efec) is a warm near-white that we found aesthetically pleasing. The tint is applied using gentle alpha blending: the model retains its original colors, while the background is uniformly colored. Hair pixels receive a gradual transition.

Two automatic crops:

Full Body (1060×1280 ratio) — the full model centered with 4% margin above and below.
Knee-cropped (1060×1280 ratio) — top of model to 65% of model height, for editorial close-up use.

Replicate — platform for open-source AI models via API. Used for rembg (background removal).
Sharp — open-source Node.js image processing for tint blending, texture, sharpness and cropping.