Navratri 2026 runs from 11 October to 19 October, with Dussehra on 20 October. That is ten nights of photos, and nobody owns ten outfits. This page gives you one complete prompt for every night: a different colour, a different fabric and a different setting each time, all built around your own face. Upload one clear selfie, paste the prompt for that day into ChatGPT or Google Gemini, and post it the same evening.
Example result — your photo will look different
How to use this page
Bookmark it and come back each night rather than making all ten at once — the results look more natural on a feed when the selfie you upload is a recent one. Every prompt below is complete on its own. The face-lock line, the framing rules and the finish are already written in, so you copy the whole block without editing anything. If a result is off, reply to the model with one short correction instead of pasting the prompt again.
Colours here follow the usual nine-night sequence loosely rather than strictly. If your family or your garba group follows a fixed colour list, swap the colour word in the prompt and leave every other line exactly as it is.
Day 1 (11 October): White and silver mirror work
Opening night, and the easiest look to get right. White with silver mirror work photographs cleanly under string lights, and it is forgiving if your selfie was taken indoors.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a waist-up three-quarter portrait of me in a white and silver mirror-work chaniya choli with a soft sheer dupatta, oxidised silver jhumkas, a maang tikka and stacked bangles. Background: a garba pandal at night with warm hanging string lights and marigold garlands, kept fully out of focus. Lighting: warm string-light glow with a soft key on the face and a golden rim light on the hair. Face and body turned toward the camera, whole head visible with space above the hair, no side profile, no extreme close-up. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch, polished natural skin. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Day 2 (12 October): Deep red bandhani with gota patti
Red works on every skin tone, which is why this is the prompt most people end up reusing outside Navratri too — for Karwa Chauth and for wedding-season photos.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a waist-up three-quarter portrait of me in a deep red bandhani chaniya choli with a gold gota-patti border, a light red dupatta over one shoulder, gold jhumkas and gold bangles. Background: a decorated garba courtyard at night with warm lanterns and marigold strings, kept fully out of focus. Lighting: warm lantern light with a soft key and a gold rim light. One hand adjusting the dupatta at the shoulder, the other at the waist, chin up, warm confident smile toward the camera, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Day 3 (13 October): Royal blue with oxidised silver
Oxidised jewellery only appears if you name it. Ask for gold and you get gold; ask for silver and you often still get gold. The words that work are heavy oxidised silver, chunky, and they are already in the prompt.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a waist-up three-quarter portrait of me in a royal blue chaniya choli with silver thread embroidery and a contrast dupatta, wearing heavy oxidised silver jewellery: chunky oxidised bangles, a layered oxidised necklace and a small nose ring. Background: a whitewashed haveli wall at dusk with warm fairy lights behind me, kept fully out of focus. Lighting: soft blue dusk light with a warm fairy-light rim. Leaning lightly against the wall, arms loosely crossed, head tilted slightly, calm half smile toward the camera, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Day 4 (14 October): Mustard yellow with marigolds
Golden hour light against a marigold wall is the most forgiving setup on this page. If your selfie is slightly soft or was shot at night, use this one — the warm haze hides a lot.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a waist-up three-quarter portrait of me in a mustard yellow chaniya choli with mirror work and a bright orange dupatta, gold temple jewellery and a maang tikka. Background: a wall of hanging marigold garlands in warm late-afternoon light, kept gently out of focus. Lighting: soft golden hour sunlight from the side with a warm glow on the skin, no harsh shadows. Mid-step walking toward the camera, one hand lifting the lehenga hem, the other reaching up toward the flowers, laughing toward the lens, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Day 5 (15 October): Emerald green in a temple corridor
The quietest look of the nine. Diya light comes from below and warms the face, so this one also works as the still frame for a short video.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a waist-up three-quarter portrait of me in an emerald green chaniya choli with a gold zari border and a sheer green dupatta, gold jhumkas and a gold kamarbandh. Background: an old stone temple corridor lined with rows of lit clay diya lamps, kept fully out of focus. Lighting: warm low diya light from the front and a soft rim light on the hair, deep shadows in the corridor behind. Standing with weight on one hip, both hands gathering the dupatta in front, serene expression toward the camera, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch, rich cinematic colour grade. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Day 6 (16 October): Black and rust, heavy oxidised
The traditional Gujarati look, and the most copied prompt on this site. A dark base is what lets the mirror work catch the light — swap it for a bright colour and the embroidery disappears.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a knee-up three-quarter shot of me in a black chaniya choli with rust orange and copper mirror-work embroidery, a rust tasselled dupatta, an oxidised silver coin belt, heavy oxidised jewellery and a large nose ring. Background: an open garba ground at night with string lights and marigold garlands melted into warm golden bokeh, no other people anywhere in the frame. Lighting: warm festive night light with a colourful rim light on the hair and shoulders. Hands raised in a garba clap above chest height, one knee bent mid-beat, bright open smile toward the camera, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Day 7 (17 October): Boys — pink kurta with an orange jacket
Male portraits fail differently: the model redraws the jawline and the hairline while keeping the eyes. This prompt locks both explicitly, which is the only reliable fix.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, jawline, hairline, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a waist-up three-quarter portrait of me in a pink silk kurta with an orange Nehru jacket and churidar, a printed stole over one shoulder, a beaded mala and an oxidised kada. Background: a decorated terrace at night with warm string lights and distant city lights, kept fully out of focus. Lighting: warm string-light key with a soft rim light along the jaw and shoulder. One hand in the pocket, the other adjusting the jacket collar, chin up, relaxed confident smile toward the camera, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Day 8 (18 October): Purple velvet with kundan
Ashtami, so the heaviest outfit of the set. Velvet needs the words soft velvet sheen next to a warm light source, otherwise the fabric renders flat and plastic-looking.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a waist-up three-quarter portrait of me in a deep purple velvet lehenga choli with a soft velvet sheen, heavy kundan and pearl jewellery, and a gold-bordered dupatta draped over one shoulder. Background: a palace-style hall with gold pillars and warm chandelier light, kept fully out of focus. Lighting: warm chandelier light from above with a soft fill on the face, the velvet catching a gentle highlight along the folds. One hand lifting the dupatta away from the shoulder, the other at the waist, poised regal expression toward the camera, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Day 9 (19 October): Peacock teal, full twirl
The last night deserves movement. Ask for motion in the fabric and the hair but not in the face, or the whole frame comes back blurred.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a knee-up three-quarter shot of me in a peacock teal chaniya choli with dense mirror work and a gold-green dupatta, layered oxidised silver jewellery and stacked bangles, caught mid-twirl with the lehenga flaring out wide, both arms lifted gracefully, hair and dupatta in motion. Background: a garba pandal at night with warm hanging lights and marigold garlands, kept fully out of focus, no other people in the frame. Lighting: warm festive night light with a golden rim light on the hair and the flying fabric. Head turned toward the camera with a wide joyful smile, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, natural motion blur only in the fabric, magazine-quality retouch. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Dussehra (20 October): Ivory and gold silk saree
The simplest prompt here and usually the cleanest result. It also carries into Diwali week with only the background line changed.
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Using my uploaded photo, keep my face, facial features, skin tone and hairstyle exactly the same – 100% identity match, no change to the face at all. Create a waist-up three-quarter portrait of me in an ivory and gold silk saree with a wide gold zari border, antique gold temple jewellery and a small bindi. Background: a sunlit open courtyard at golden hour with soft warm haze and blurred trees, kept fully out of focus, with no idol, effigy or statue anywhere in the frame. Lighting: warm golden hour sunlight from behind creating a soft halo on the hair, with gentle fill on the face. Standing tall with good posture, one hand holding the saree pallu at the shoulder, calm confident expression toward the camera, whole head visible, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch. Vertical 4:5 format. No text, no watermark.
Results vary a little every time — run it 2–3 times and keep the best one.
Five things that decide the result
Your selfie does most of the work. Bright, sharp, front-facing, taken near a window. A dark upload is behind almost every complaint that the face changed.
Keep the shot waist-up or knee-up. In a full-body result the mirror work turns into grey speckles on a phone screen.
Tie your hair back in the reference photo if you want the jhumkas to show.
Run each prompt two or three times. The pose and the folds change every run and the third one is usually the best.
Correct with one short line, never by pasting a rewritten prompt. “Make the choli darker” works; a new paragraph resets everything including the face.
FAQs
Do I have to follow the colour for each day?
No. The nine-night colour list varies between regions and between publishers, and it shifts with the weekday each year. Treat the colours here as a set of ten looks and swap the colour word if your group follows a different list.
Can I make all ten in one sitting?
You can, and free tiers usually allow it across a day or two. They tend to look better spread out, though, because a photo posted the same evening it was made matches what everyone else is posting that night.
Will the same face come out every time?
Close, if you upload the same selfie each day and keep the face-lock line at the start. Using a different selfie every day is what makes a ten-photo set look like ten different people.
Is it free?
Yes. Every prompt on this page is free to copy, and both ChatGPT and Google Gemini have free tiers that accept photo uploads. The only limit is how many images the free tier gives you per day.
Related looks
For the full write-up on the traditional black and rust outfit see the chaniya choli prompts, and the dance-floor version lives in the garba prompt. To build a look that is not on this list, use the Prompt Builder, and the prompt guide explains what every line in these prompts is doing.
Want a different look?Build your own prompt — pick the outfit, background and lighting, then copy it in one tap.
AI Photo Editing
AI Photo Editing
AI Photo Editing
AI Photo Editing
AI Photo Editing
AI Photo Editing