When the Food Composition Differs from the Reference Image
In QUEL, results are created using prompts along with reference images. In this case, the image reference serves as a guide for the overall mood, angle, and direction, while the composition or appearance of the food is greatly influenced by the text prompt. Therefore, even if a reference is used, if the prompt is simply written as "hamburger," "pasta," or "lunch box," the system cannot accurately determine the specific shape the user wants. As a result, food with a different composition from the reference may be generated. This difference occurs more frequently in the following cases: - When it is difficult to clearly distinguish what food it is from the reference image alone - When the composition of the food in the reference and the description in the prompt differ In such cases, the most reliable solution is to directly describe the composition and shape of the food in text.
Examples of Poor Prompt Writing
1) When only the food name is written without a reference
“Chicken burger”
“One hamburger”
If you only write the name like this,
important elements such as patty thickness, sauce color, and ingredient composition are not defined,
so any style of burger can be broadly interpreted.
2) When a chicken burger photo is used as a reference and the text is written simply
Reference: Chicken burger photo
Prompt example:
“A hamburger is on the table.”
“A hamburger is placed in a white studio.”
In this case, the text does not include the information 'chicken'.
From the system's perspective, it is recognized as information closer to a “general hamburger”
rather than a “specific chicken burger”,
and a different type of burger may be generated accordingly.
Examples of Correct Prompt Writing
Using reference + text description together can yield much more stable results.
The key is to specifically describe the ingredients, color, thickness, and shape.
Example 1 ) Sesame bun, thickly fried crispy chicken patty, lettuce, orange spicy sauce mixed with roasted vegetables
Example 2 ) Sesame bun, mayonnaise sauce, lettuce, thin and wide fried crispy chicken patty, pink roast ham slice
If you write to this level,
the system can generate an image based on the composition and shape defined in the text
while referring to the reference image.
Summary of Prompt Writing Criteria
When using a reference image, please include the following items in text as much as possible.
Composition ingredients
Bun, sauce, patty/noodles, vegetables, meat, etc.
Color of ingredients
Orange sauce, pink ham, etc.
Texture and thickness
Thick patty, thin and wide patty, glossy noodles, etc.
Arrangement and organization
Noodles neatly arranged, toppings neatly placed on top, etc.
If you write the reference image and text description together in detail like this,
you can generate food images in QUELSUITE much more stably and closer to the desired direction.