Skip to main content

Generation instruction

To generate the desired image of your product using our AI model (hereinafter referred to as AI), the following parameters are mandatory:

  1. A photograph of the item (clothing, footwear, accessories, etc.) on a neutral background (white, light, grey - ensuring the item is clearly visible).
  2. A textual description (prompt) of the desired generation - describing the person wearing the item and the background against which the person will be generated.

Requirements for the item photograph: 

  • The item photograph must be of high quality on a neutral (white) background. AI is trained on such data, enabling higher-quality generations. 
  • Only one photograph of the original item is used for generation. If necessary, such as for generating a suit comprising a top and pants, you may merge their photographs into one image for input to AI. 
  • A photograph of the item on person can be used, but this significantly reduces generation quality. Additionally, AI will generate an image based on a similar person (changing the prompt in this case is unlikely to help). Therefore, if another photograph of the item is unavailable, it is preferable to mask or crop the person's face in the original photograph. 
  • Generation occurs according to the perspective of the original item photograph. You can upload front, back, or side views of the item and specify the model's position for generation (front, back, side). If you upload a front view and request a back view, AI can accommodate this request, though the resulting back view may differ from the original item. This method may work well for basic items like solid-color shirts and pants.

Requirements for the textual description (prompt): 

  • The minimum prompt should include: a description of the person in the photograph (gender, race), a description of what they are wearing (including the item for which the image was input), and a description of the background. 
  • The description of the person can be detailed with any desired parameters: hair color, skin tone, eye color, height, age, and body type (slim, Plus size, etc.). 
  • If you request a specific ethnicity (Arabic, Asian, etc.), AI will adjust eye and hair colors accordingly. 
  • You may also specify additional clothing or accessories for the model. For instance, if you upload a photo of a t-shirt, you can specify that the model should wear the t-shirt, jeans, sneakers, and carry a sports bag. If not specified, AI will autonomously select additional clothing and possibly accessories. 
  • In the prompt, you can indicate the desired format of the generation (full-length photo, portrait, perspective, etc.). 
  • Detailing the original item description is not mandatory. However, it should be understood that when uploading, for example, a dress, AI may not understand its length, so it is advisable to specify the dress model - mini, midi, maxi, and you can also specify the length by other means (knee-length, calf-length, etc.). Additional details in the prompt can assist AI in generating more accurately, such as specifying thin shoulder straps for a dress and etc. 
  • The background of the photo can be described broadly (studio, Paris, Santorini, etc.) or in detail as needed (including specific items or locations like the Eiffel Tower in Paris). You can also specify the time of day and desired lighting (dawn, daytime, evening, etc.).

 

Additional parameters during generation. 

During generation, the bot allows adjustment of several parameters (click on +3 more after the prompt; these parameters are optional):

  • Aspect ratio - size of the generated images (e.g., 7x9, 19x13, etc.).
  • Negative prompt - specify anything you do not want to see in the generation. For example, if you want to generate a person on a street without cars, include "cars" in the negative prompt. You can also state this in the generation prompt (e.g., street scene without cars in the background).
  • Seed - fixes the generation process for replication, provided the same photo and prompt are used. Useful if you are satisfied with the generated item but wish to change the model (e.g., hair color) or generate a different background.

Here you can find examples of prompts:

  • Full-length photo of beautiful arabic woman in black oversized blazer and black flared pants, no pockets, dark hair, brown eyes, makeup, high heels, golden earrings, detailed hands, posing, beautiful modern office building background, ultra realistic photo
  • Full - length photo of 5 years old european girl in dress , brown ponytails hair, smiling slightly, posing, summer park background, ultra realistic photo, studio light
  • Portrait photo of Beautiful caucasian woman with blonde wavy hair, detailed brown eyes, in t-shirt, jeans, white sneakers, fashion posing, Santorini background
  • Profile photo of beautiful 14 years old kazakh girl in white baseball cap with no logo, smiling slightly, uniform studio background, ultra realistic photo, uniform studio light
Was this article helpful?