 (3).png)
4 months ago
9
ARTICLE AD BOX
Google's Gemini Flash Image AKA Nano Banana has social media on edge over the last few days with users taking advantage of the model's advanced capabilities to power a new vintage Saree trend among other things. While most users just copy and paste prompts from various sources for transforming their images into different effects, one can also improve the output from Gemini by following a few basic principles when giving prompts to the model.
Here's what you need to keep in mind when prompting Gemini Nano Banana model:
Tips to keep in mind when prompting Nano Banana:
1) Start with a simple formula:
Start your prompts with a simple three part structure that includes the who, what and where of the image you want to generate before adding further details so that the model doesn't have to guess. If you start by telling the model the specific style, lighting, mood, camera angles all at once, it could mis-read what is important and what is optional.
To explain this point, Google gives an example of: 'Create an image of a cat napping in a sunbeam on a windowsill'.
Essentially your first prompt should have the ‘subject’, ‘action’ and ‘scene or background’.
2) Be specific:
Try to avoid generic words in your prompts by adding details like age, attire, facial expression, fabric texture, setting, lighting, accessories.
So for instance, instead of using a prompt like “a woman in a red dress…”, replace it with “a young woman in a deep-crimson silk saree with a gold zari border, standing under a banyan tree at dusk.”
Generative AI models interpolate the missing details from billions of possiblities, which means that the defaults selected by the model would likely clash with your vision of the image.
3) Think about the style:
While transforming your picture into a different style, you need to careful not just while chossing the scenario where you want your picture to be transported like Polaroid, 90s Bollywood posters or retro-grainy, but also about the general mood you want to evoke with the image like dreamy, dramatic, nostalgic, glamorous or romantic.
You can also give the AI model lighting cues like image in golden hour, neon-lit night, soft backlight and so on. All of these can change the model of the image drastically.
Sample prompt: “A dreamy, grainy 35 mm film portrait of a dancer twirling in a dimly lit jazz club.”
4) Composition:
The AI model will decide where each element in the image will be placed unless you give it specific details. Telling the model details camera angle, framing (close up or full body portrait) and depth of field can help improve the composition of the image while elevating visual impact.
Sample prompt: “In the foreground, a woman draped in a flowing blue saree; in the blurred background, a bustling Diwali market.”
5) Quality and aspect ratio:
Nano Banana usually delivers the image in a square crop and with the added pressure on Google's servers, the model has started giving images in low resolution many times. To rectify that try to state the resolution of image, like 4K or high-res along with the aspect ratio you need it in like 16:9 for banners, 3:4 for portraits and 1:1 for Instagram.
In case you still get an issue with the image after following all of these instructions, you can always ask Gemini to take specific changes in the images.
Pro Tip: Use Qwen's Image Edit feature in order to remove the Gemini logo while posting on social media along with improving the details in the image.
“Remove the Gemini logo at the bottom right and enhance the resolution while keeping all other elements exactly the same.”
You can also ask Qwen to change the aspect ratio of the image generated by Gemini.
Prompts for Gemini Saree trend:
Prompt 1:
Convert the uploaded image into a stunning 4K HD portrait. The subject should have long, dark, wavy hair cascading over her shoulders. She should be wearing a translucent, elegant red saree draped over one shoulder, which reveals a fitted blouse underneath. White flowers should be tucked behind her right ear. She is looking slightly to her right, with a soft, serene expression. I want her face to remain exactly as it appears in the uploaded image without any alterations. The background should feature a plain, warm-toned wall, illuminated by a warm light source from the right, creating a distinct, soft-edged shadow of her profile and hair on the wall behind her. The overall mood should be retro and artistic.
Prompt 2:
Create a retro, vintage-inspired image—grainy yet bright—based on the uploaded picture. The girl should be draped in a perfect purple chiffon, Pinterest-style aesthetic saree. The vibe must capture the essence of a '90s movie 'baddie,' with dark brown, silky hair and a small flower tucked visibly into her hair, enhanced by a windy, romantic atmosphere. She is standing against an old wooden door, where deep shadows and dramatic contrasts add mystery and artistry to the scene, creating a moody yet enchanting cinematic effect. Her pose should suggest that she is adjusting her hair.
Prompt 3:
Convert the uploaded picture into a retro, vintage, and grainy-but-bright image. Keep the facial features the same. The subject should be draped in a solid-coloured Banarsi saree with a Pinterest-retro aesthetic, giving it the feel of a '90s movie. Give her dark brown, silky, and shiny hair with a small flower visibly tucked into it. The girl is standing in front of a white wall with a soft, artistic, and moody atmosphere. Use a golden light source to create a soft glow on her face and a dramatic shadow on the wall behind her. The lighting should have the warm, golden tones of a sunset or 'golden hour' glow. The background should be minimalist and slightly textured. The expression on her face should be moody and calm, yet happy. The final image should be in HD quality, with the same face, but you can change the hair colour and hairstyle.
English (US) ·