A Chhath Puja AI photo turns one selfie into a portrait of you on a river ghat, in a red saree or a white kurta, with a bamboo soop of fruit and lit diyas beside you. On videosi you skip the prompt: upload one photo, pick a setting and get one vertical 2:3 still image for 1 credit.
This guide goes deeper than the FAQ on the trend page. It covers how to describe the soop and the offerings, what men and women wear, sunrise or dusk light, why AI tends to add a glowing figure and why we never draw one, how to keep your face, and how to post the result. Chhath Puja 2026 runs from 13 to 16 November, and the main day is Sunday 15 November.
Key takeaways
- The picture is built from objects people really carry to the ghat: a round bamboo soop, bananas and oranges, sugarcane, thekua and small clay diyas.
- The result is one vertical 2:3 still image made with GPT Image 2.5 from OpenAI, not a video.
- videosi draws no deity, idol or sun god figure, so the scene shows the offerings, the river and a low sun only.
- Pick ghat steps at sunrise or a diya-lit ghat at dusk, and use one sharp front-facing selfie to keep the face yours.
- Each image costs 1 credit, and a failed render returns the credit automatically.
In this guide
- How do you describe the soop and the offerings?
- What do women and men wear in a Chhath photo?
- Sunrise or dusk: which setting is better?
- Why does AI add a glowing figure to the scene?
- How do you keep the face looking like you?
- Which selfie works best?
- How does this compare with Gemini and ChatGPT?
- How do you post it on Instagram and WhatsApp?
- When is Chhath Puja in 2026?
- What does it cost and what if it fails?
How do you describe the soop and the offerings?
This is where short prompts fail. If you only write "Chhath Puja", an image model usually returns a wide crowd scene where you are a tiny figure, and the offerings turn into a vague basket of fruit. The soop is a round, flat bamboo winnowing tray, and it is the centre of the picture, so it needs its own description.
The videosi recipe lists the objects one by one: the bamboo soop holding bananas and oranges, a few sticks of sugarcane, thekua sweets and small lit clay diyas. Each item has a size and a place, and the soop sits beside you on the stone step or at waist height in your hands. If you write your own prompt elsewhere, name every object and say where it sits. A model that has to guess will invent steel pots, flower garlands or plastic trays.
Behind you, softly out of focus, adult devotees stand in the river in orange and red clothes with their hands raised. Keeping them blurred matters, because sharp faces in the background compete with yours. You can see the finished look on the Chhath Puja AI photo page.
What do women and men wear in a Chhath photo?
For women, the outfit is a plain silk or cotton saree in bright orange-red or red with a gold border, a matching blouse, the pallu over one shoulder, glass bangles, small gold earrings and a small bindi, with the hair neatly tied. Plain is the right word: a heavily embroidered bridal saree looks like a wedding, not a ghat.
Many searches are for a boy version of the prompt, so the photo handles men too. If your selfie shows a man, the outfit becomes a plain white or pale yellow kurta with a white dhoti or pyjama and a cotton stole over one shoulder. There is no jewellery and no bindi. Glasses, facial hair and a head covering such as a hijab are kept from your photo, because only the clothes and the setting change.
Sunrise or dusk: which setting is better?
You get two settings, and they suit different posts.
Ghat steps at sunrise puts you seated on wide stone steps, waist-up and slightly angled, with one hand on the soop. Rows of brass plates with fruit and diyas run down the steps, and a low orange sun sits over a hazy river. It is the cleaner pick for a bright festival portrait and a profile picture.
The diya-lit ghat at dusk has you side-on and smiling, holding a soop of small lit diyas at waist height, with circles of diyas along the steps and a pale pink sky over the river. It is warmer and moodier, and it suits an evening Story. If you cannot decide, make one of each. At 1 credit per image the pair is cheap, and they look quite different side by side.
Why does AI add a glowing figure to the scene?
Image models learn from captions, and many captions of Chhath mention the sun god or Chhathi Maiya. As a result a model often adds a glowing human shape in the sunlight, a statue-like form on the steps or a haloed figure above the water, even when you never asked. Prompts that name the deities make this much more likely.
videosi does not draw deities, idols or any religious figure. The prompt leaves those names out and says plainly that no figure of any kind may appear, so the sun is only a low orange disc over the river. If you see a figure in a result from another tool, that is the model guessing from its training data. Leaving the names out and asking for offerings only is the most reliable fix.
How do you keep the face looking like you?
The most common complaint about AI festival photos is that you become a more glamorous stranger. The fix is to put the identity rule first and make it strict. In the videosi prompt, your face, age, skin tone, features, build and hair all come from your photo, and only the clothes and the setting change. Short copied prompts often skip this, so the model improves your face without being asked.
Warm sunset light and new clothes can still shift small details, and likeness is not guaranteed. After the result arrives, compare it with the original in good light. If it looks like a relative and not you, try a different selfie before changing anything else.
Which selfie works best?
Pick a front-facing photo in soft daylight, with the whole face visible and nobody else in the frame. A shoulders-up or waist-up shot is ideal, because the result is also a waist-up portrait. Avoid sunglasses, masks, hands near the face and heavy beauty filters, since a filtered selfie gives the model a different face to copy.
Group photos do not work, because the model cannot know which person you mean. Crop the photo down to the one face you want. A relaxed or lightly smiling expression gives the model the most room, and a stiff passport look can fight with the happy, calm expression the prompt asks for.
How does this compare with Gemini and ChatGPT?
Most Chhath Puja AI photos in 2025 came from long prompts pasted into Gemini or ChatGPT, often with a separate boy or girl version. You can still do that, and it can work. The usual problems are the ones above: the face drifts, you shrink into a crowd, the soop is vague and a figure appears in the light.
With videosi you do not write or paste anything. The identity lock, the soop and diya details and the men's outfit are built in, and you choose only the setting. The trade-off is that it is not free: you pay 1 credit per image, while a copied prompt can be tried in Gemini itself under its own limits. videosi is independent and not affiliated with Google or OpenAI.
How do you post it on Instagram and WhatsApp?
The result is a 2:3 vertical image, which sits well on an Instagram feed post and fits a Story or a WhatsApp status with only slight cropping. For a Story, leave room at the top and bottom, since you and the soop sit in the middle of the frame.
On WhatsApp, send it as a document if you want to keep full quality, because a photo is compressed. A short caption with the festival name and a line such as "Happy Chhath Puja" is enough. The image has no sound, so add a Chhath geet or your own audio if you turn it into a reel. If you share it publicly, say that it is AI made, which helps friends who expect a real ghat photo.
When is Chhath Puja in 2026?
Chhath Puja 2026 runs from 13 to 16 November, with the main day, when the evening offering is made, on Sunday 15 November. The festival follows the lunar calendar, so the dates move each year. It falls in the days after Diwali, mainly in Bihar, Jharkhand, eastern Uttar Pradesh and Nepal's Terai.
Make your photo a few days before 13 November. That leaves time to try the other setting, fix a bad selfie and post on the first day, when feeds are busiest. The trend page stays live all year, so you can also keep it for next season.
What does it cost and what if it fails?
Each image costs 1 credit. Credit packs start at $6.99 for 10 credits, and credits never expire, so you can keep them for the next festival. If a render fails or is blocked, the credit is returned automatically, so you only spend credits on images you actually receive.
Because the result is a still image and not a video, there is no long wait, and it is usually ready in about a minute. It is made for one adult per image. For a couple or family, make one image each with the same setting and post them together, and only use photos of adults who have agreed.
