Best Photos for a Gangnam Style AI Video: A Practical Upload Guide

Start with an ordinary, well-lit portrait. You should be able to distinguish the eyes, mouth, hairline and overall face shape without zooming far in. A simple background helps you judge those details, but you do not need a studio backdrop or matching backgrounds for all three people.

Compare two candidate photos at the same size. If one has a clearer face but a less interesting location, use the clearer face. The template supplies the moving scene; the upload supplies the person. A tiny face inside a beautiful holiday landscape gives the generator less useful identity information.

• Use one person per file. Crop a group photograph into separate portraits only if each resulting face remains clear.
• Avoid sunglasses, hands over the face, heavy shadows and motion blur. Keep hair visible rather than cropping tightly around the eyes and mouth.
• Choose a photo whose clothes and hairstyle you want represented. The reference is used for appearance beyond just the face.
• Skip collages, decorative stickers and strong beauty filters that hide the original facial details.

A head-and-shoulders or upper-body picture is a practical starting point: it leaves room for hair and clothing while keeping the face readable. Do not stretch a narrow crop to make it wider. Distortion becomes part of the reference the generator sees.

The upload order is not left to right across the finished picture. Decide who should take the first solo, place that photo in Middle, then fill Right and Left. Check the thumbnails once more before submitting. A correct role assignment is easier to establish before generation than to diagnose from a fast camera transition afterward.

If you want the same person in all three roles, use the dedicated one-selfie mode. For animals, follow the pet photo guide.

The uploader accepts JPEG, PNG and WebP, up to 10 MB per photo. Export an unsupported phone image into one of those formats; changing only the filename does not convert it. If a file is too large, export a smaller copy while checking that the facial details remain visible.

Each side must be 301–5,999 pixels, with width divided by height strictly above 0.4 and below 2.5. Crop extremely narrow or wide images while keeping the face clear.

Use photos you own or have permission to use, and confirm consent. Check your current account or guest wallet balance; follow any sign-in prompt shown. Review the thumbnails before generating to catch the wrong file, an accidental duplicate or an obscured face.

Open the Gangnam Style studio, choose 480p or 720p and check the displayed credit cost. The current output is a landscape 16:9 MP4 based on a 15-second reference, with its original soundtrack added automatically. The generated clip can be slightly shorter.

After delivery, inspect the opening group, all three solos and the closing group. If only one person looks unclear, consider replacing that reference rather than changing every photo at once. Higher resolution does not guarantee a better identity match. Each new generation has its own cost; the cost and fixes guide explains the difference between a failed task and a completed result you want to improve.