Saree photos are back on every feed this month, and the ones getting saved are not studio shoots — they are selfies run through a prompt. Five looks below, each a complete prompt you copy as it is: a filmy heroine portrait, a royal Kanjeevaram, a retro chiffon, a bright festival yellow and a moody twilight velvet. You do not need the outfit, the location or an editing app.
Example result — tumhari photo alag aayegi
What makes a saree photo work
Three things decide it, and all five prompts below hold them. First, the fabric has to read as fabric — silk, chiffon and velvet each catch light differently, so the prompt names the fabric and the light together rather than just saying “saree”. Second, the background stays out of focus. A sharp palace pillar competes with the drape; a blurred one frames it. Third, the frame stays waist-up. In a full-body shot on a phone screen the zari work turns into a thin gold line and the whole point is lost.
Colour matters more here than in most looks. Deep reds, royal blues and emerald greens hold their richness after Instagram compresses the image. Pale pastels flatten out, which is why the one pastel look below is built around soft window light instead of colour.
The prompts
Prompt 1: Classic filmy heroine, hill station sunrise
The timeless one — red chiffon, fog and sunrise light. Start here if you are only making one photo.
AI example — aapka result alag ho sakta hai
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Edit my uploaded photo. If I attached more than one photo of myself, use all of them together to get my face right. Do not redraw my face – keep it exactly as in the original: the same face shape and width, the same cheeks, the same jawline and chin, the same nose shape and width, the same eye shape and spacing, the same eyebrows, the same lip shape, the same skin tone and the same skin texture. Do not slim my face, do not sharpen the jaw, do not enlarge the eyes, do not smooth or airbrush the skin, do not change my age and do not make my features more symmetrical. Change ONLY the clothing, the background and the lighting described below, and keep my head at the same angle as in the photo. Create a waist-up three-quarter portrait of me in a classic red georgette saree with a thin gold border and a matching sleeveless blouse, wearing simple oxidised silver jhumkas and a small bindi. Background: a rustic wooden balcony overlooking a foggy hill station at sunrise, kept fully out of focus. Lighting: soft golden sunrise light from the side with a warm halo on the hair, low contrast and nostalgic. One hand resting on the railing, the other holding the pallu, head turned toward the camera with a soft smile. 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, natural skin. Vertical 4:5 format, high resolution. No text, no watermark.
Chehra same rehne ki guarantee nahi hai. Result ko apni reference photo se compare karo aur zaroorat ho to correction do.
Prompt 2: Royal Kanjeevaram in a palace courtyard
The grand version, built for wedding season and festival posts. Heavy zari needs warm lamp light to show up, and this prompt asks for exactly that.
AI example — aapka result alag ho sakta hai
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Edit my uploaded photo. If I attached more than one photo of myself, use all of them together to get my face right. Do not redraw my face – keep it exactly as in the original: the same face shape and width, the same cheeks, the same jawline and chin, the same nose shape and width, the same eye shape and spacing, the same eyebrows, the same lip shape, the same skin tone and the same skin texture. Do not slim my face, do not sharpen the jaw, do not enlarge the eyes, do not smooth or airbrush the skin, do not change my age and do not make my features more symmetrical. Change ONLY the clothing, the background and the lighting described below, and keep my head at the same angle as in the photo. Create a waist-up three-quarter portrait of me draped in a rich royal blue Kanjeevaram silk saree with heavy gold zari motifs, wearing an elaborate gold choker, matching bangles and a maang tikka. Background: an ancient palace courtyard with carved marble pillars and warm brass hanging lamps at twilight, completely out of focus. Lighting: warm lamp light as the key with a gold rim light on the hair and shoulder, deep shadows behind. Standing tall beside a pillar, one hand holding the pallu at the shoulder, poised regal expression toward the camera. Whole head visible with space above the hair, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch, rich silk texture. Vertical 4:5 format, high resolution. No text, no watermark.
Chehra same rehne ki guarantee nahi hai. Result ko apni reference photo se compare karo aur zaroorat ho to correction do.
Prompt 3: Retro chiffon, sunlit vintage room
The softest of the five. This is the only look where film grain helps — on the other four it just makes the fabric look dirty.
AI example — aapka result alag ho sakta hai
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Edit my uploaded photo. If I attached more than one photo of myself, use all of them together to get my face right. Do not redraw my face – keep it exactly as in the original: the same face shape and width, the same cheeks, the same jawline and chin, the same nose shape and width, the same eye shape and spacing, the same eyebrows, the same lip shape, the same skin tone and the same skin texture. Do not slim my face, do not sharpen the jaw, do not enlarge the eyes, do not smooth or airbrush the skin, do not change my age and do not make my features more symmetrical. Change ONLY the clothing, the background and the lighting described below, and keep my head at the same angle as in the photo. Create a waist-up three-quarter portrait of me in a sheer pastel pink chiffon saree with tiny floral prints and light jewellery, seated on a cane chair in a sunlit vintage room with indoor plants and old wooden furniture, all kept out of focus. Lighting: soft afternoon window light from one side casting gentle shadows, with a light 1970s film grain and slightly faded highlights. Seated with the torso turned three-quarters, one arm on the chair, looking into the lens with a calm half smile. Whole head visible with space above the hair, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch that keeps natural skin texture. Vertical 4:5 format, high resolution. No text, no watermark.
Chehra same rehne ki guarantee nahi hai. Result ko apni reference photo se compare karo aur zaroorat ho to correction do.
Prompt 4: Festival yellow, marigolds and diyas
The brightest one, and the most forgiving if your selfie was shot indoors or came out slightly dull.
AI example — aapka result alag ho sakta hai
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Edit my uploaded photo. If I attached more than one photo of myself, use all of them together to get my face right. Do not redraw my face – keep it exactly as in the original: the same face shape and width, the same cheeks, the same jawline and chin, the same nose shape and width, the same eye shape and spacing, the same eyebrows, the same lip shape, the same skin tone and the same skin texture. Do not slim my face, do not sharpen the jaw, do not enlarge the eyes, do not smooth or airbrush the skin, do not change my age and do not make my features more symmetrical. Change ONLY the clothing, the background and the lighting described below, and keep my head at the same angle as in the photo. Create a waist-up three-quarter portrait of me in a bright yellow organza saree with embroidered floral borders, gold temple jewellery and a small bindi. Background: a courtyard strung with marigold flowers and glowing earthen diyas in the late afternoon, kept softly out of focus. Lighting: warm bright daylight from the side that makes the colours pop, no harsh shadows. Mid-movement with one hand lifting the pallu, shoulders open, bright smile toward the camera. Whole head visible with space above the hair, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch, vivid warm colour grade. Vertical 4:5 format, high resolution. No text, no watermark.
Chehra same rehne ki guarantee nahi hai. Result ko apni reference photo se compare karo aur zaroorat ho to correction do.
Prompt 5: Twilight velvet by an arched window
The moody option. Cool window light against a warm indoor lamp is what gives this one its depth.
AI example — aapka result alag ho sakta hai
✦ AI PROMPTCopy → paste in ChatGPT / Gemini
Edit my uploaded photo. If I attached more than one photo of myself, use all of them together to get my face right. Do not redraw my face – keep it exactly as in the original: the same face shape and width, the same cheeks, the same jawline and chin, the same nose shape and width, the same eye shape and spacing, the same eyebrows, the same lip shape, the same skin tone and the same skin texture. Do not slim my face, do not sharpen the jaw, do not enlarge the eyes, do not smooth or airbrush the skin, do not change my age and do not make my features more symmetrical. Change ONLY the clothing, the background and the lighting described below, and keep my head at the same angle as in the photo. Create a waist-up three-quarter portrait of me in an emerald green velvet saree with an embroidered blouse and subtle gold jewellery, standing near a large arched window at night. Background: the window and a warm chandelier glow, fully out of focus, blue hour light outside. Lighting: cool blue window light mixed with a warm indoor key on the face, elegant and calm, deep shadows around. One hand resting on the window frame, head turned toward the camera with a composed expression. Whole head visible with space above the hair, no side profile. Finish it like a professional model shoot: 85mm lens look, shallow depth of field, magazine-quality retouch, rich velvet texture. Vertical 4:5 format, high resolution. No text, no watermark.
Chehra same rehne ki guarantee nahi hai. Result ko apni reference photo se compare karo aur zaroorat ho to correction do.
How to use them
Open ChatGPT or Google Gemini on your phone. Both are free and both take photo uploads.
Upload one bright, sharp, front-facing selfie. Daylight near a window is the easiest way to get one.
Paste the whole prompt in the same message as the photo and send.
If something is off, reply with one short correction such as “keep my face exactly like the uploaded photo”. Do not paste the prompt again — that resets everything including the face.
Run the same prompt two or three times. The drape and the pose change every run and the third one is usually the best.
Crop the result to 4:5 before posting with the photo resizer — it runs in your browser and nothing is uploaded.
Tips that change the result
Tie your hair back in the reference selfie if you want the jhumkas to show. Hair over the ears hides the jewellery the prompt asks for.
A dark or blurry upload is behind almost every complaint that the face changed. The selfie matters more than the prompt.
Keep one look per post. Five different sarees in one carousel reads as a prompt dump; five poses in one saree reads as a shoot.
Do not over-sharpen the downloaded image. Silk and velvet lose their softness fast, and that softness is what makes the photo look expensive.
FAQs
Is it free?
Yes. Every prompt here 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.
Will it keep my real face?
Yes, if your selfie is bright and front-facing and the identity-lock line stays at the start of the prompt. That line is the first sentence of every prompt above — do not delete it.
Can I change the saree colour?
Yes, swap the colour word and leave every other line alone. Keep the colour deep rather than pale, or the fabric loses its richness once Instagram compresses the photo.
Can I use these for Reels?
Yes. Generate three or four from the same prompt, crop to 9:16, and drop them into a slow template. The Dil Lagana Mana Tha template is built for exactly that kind of set.
