Prompt deep-dive
Prompting Cheatsheet
Quote text
Every non english prompts submitted to the application are translated in english, prior to be moderated and upscaled.
But users may want to generate text, in their own language, in the target image.
Preserve source language
To prevent text translation, surround it between double quotes like this:
Un homme qui tient un écriteau avec écrit *"Parlez-vous Francais ?"* dessus
The prompt above will be translated then upscaled like this:
a man, holding a sign that says *"parlez-vous francais ?"*, urban street background, candid street photography, natural lighting, 50mm lens, photo-realistic, documentary style
Generated image with prompt upscale
Translate source language
In case of you may want to generate text in english, surround it between simple quotes like this:
Un homme qui tient un écriteau avec écrit ‘***Parlez-vous Francais ?’*** dessus
the prompt above will be translated then upscaled like this:
a man holding a sign that says *"Do you speak French ?"*, in a busy urban street, natural lighting, in the style of street photography, close up, candid shot, portrait orientation, 50mm lens, high detail
Generated image with prompt upscale
Typography control
Capitalize text
prompt: book with text "Welcome" capitalized
Add typographic style to your text
prompt: photo of a paper with hand written text "Welcome”
Describe colors
IA generative models are very versatiles and understand well colors instructions. But actually, they are unable to respect exact colors RGB or HEX values.
Generally, we discourage usage of color specific codes like RGB or HEX values, and prefer textual guidance, even if less precise.
That said, Seelab will find the best color fit according the code you provide. But once again, don’t expect precise color, but a fidelity in tone.
Examples
Botanic Figure of a red mushroom, #FFF background
Botanic Figure of a red mushroom, 173,216,230 background
Botanic Figure of a red mushroom, dark orange background
A #FF4500 fish on black background
Auto Aspect Ratio [WIP]
When the format selector is set to Auto, Seelab will select the “best” aspect ratio according the image you want to generate.
In this mode, It is also possible to indicate the desired ratio right from the prompt, Seelab will care about generating the image in the corresponding format.
Seelab actually supports the following aspect ratios: 1:1, 3:4, 4:3, 9:16, 16:9
Examples
Black and white portrait of a man looking at a sunset
Images are generated in 3:4 format that works well with portraits
Cinematic black and white photo of a man looking at a sunset
16:9 wide screen format was chosen because of the “cinematic” keyword
Instagram story black and white photo of a man looking at a sunset
9:16 story format was chosen. Note: Flux do not consider Instagram as a monochrome place !
Updated on: 07/01/2025
Thank you!