Transmission 5 Dance Photos: Preparation and Quality Checks

Clear photos make it easier to assess an AI dance result, but they do not guarantee a recognizable face or natural motion in every frame. The homepage studio accepts new generations using credits; review the existing examples before submitting your photos.

Prepare one consenting adult per photo

Use your own photo or a photo of an adult who agreed to the intended use. Prefer a sharp, well-lit face, even lighting and a neutral standing pose. A full-body reference provides information about clothing, posture and shoes that a cropped selfie cannot show.

Avoid heavy shadows, faces hidden by hair or sunglasses, and poses where limbs merge with the background. Missing body and clothing details may be invented by AI. Do not assume a better photo eliminates that uncertainty.

Match the input to Solo or Duo

Solo uses one individual photo to replace the left dancer and keep the original partner. Duo uses two separate individual photos, one for each dancer. A combined group photo is not the input demonstrated by the published Duo example.

See the Solo and Duo comparison.

Why can a face or body change?

The model must generate new views and motion from limited appearance information. A small or obscured face, fast movement and overlapping limbs can make the result harder to keep consistent. The output may change facial features, invent clothing, distort hands or feet, or mix details between two people.

A general text prompt cannot guarantee a fix. A cleaner reference can help you evaluate the result; if the workflow allows motion controls, changing the reference or shortening a test may also be worth evaluating. The current studio uses a fixed reference dance and duration; it does not expose motion-reference or duration controls.

Inspect the complete clip

• Face: does the person remain recognizable during turns and fast movements?
• Body: do arms, legs and joints move plausibly, including during overlaps?
• Clothing: do outlines, patterns and shoes remain consistent?
• Timing: do the dancers follow the reference together without sudden jumps or flicker?
• Whole video: check transitions between frames, not just a thumbnail or selected still.

The three published tests are 10-second, 720p, vertical 9:16 videos without audio. They are individual results, not independently measured success rates or evidence that every photo will work equally well.

What can be fixed afterwards?

Cropping or editing can remove an unwanted portion, but does not establish that an incorrect face or body has been repaired. Sharpening and upscaling do not reliably reconstruct the intended anatomy. Face enhancement may change identity further. Review any edited version before sharing it.

You can watch the existing examples and read how the demonstrated workflow uses a dance reference. New generation remains unavailable; check pricing before buying credits; see current availability.

When sharing an AI edit, disclose it and use audio you are authorized to use. For questions about these examples, contact support@transmission5danceai.org.