Alors on Danse AI troubleshooting

The wrong person is on the keyboard

Cause: Photo 1 always becomes the performer on the left at the mini keyboard; Photo 2 is always the hype partner on the right. Other tools number their photos differently.

Fix: press Swap photos in the generator and generate again.

A face drifts or the two faces blend

Cause: identity drift. The model keeps two separate people while they dance, turn and lean into the camera, and the close-up of Photo 2 is the hardest moment. Similar-looking people can swap features.

Fix:

1. Use one person per photo, facing the camera, in even light.
2. Avoid sunglasses, masks, heavy filters and very small or blurry faces.
3. Use waist-up photos so the model also gets the outfit.
4. Pick photos where the two people look different: hair, clothing colour, background.
5. Use Seedance 2.5 or Seedance 2.0 instead of Mini.
6. Test a 5- or 10-second clip before generating 30 seconds.

Faces change halfway through a 30-second video

Cause: on Seedance 2.0 Mini, Fast, 2.0 and MiniMax H3, 30 seconds is two 15-second generations joined together (Mini stops at 15 seconds). Faces can shift slightly at the join.

Fix: use Seedance 2.5, which checks both photos and the reference clip first and renders 30 seconds in one pass.

My pet looks strange or too human

Cause: the reference is a two-person performance, so a pet becomes an upright character that copies human moves.

Fix: use a clear photo of the head and some of the body, one animal per photo, and Seedance 2.5. More in Alors on Danse AI with pets.

The downloaded video has no sound

Cause: the model renders silent video, and the finishing step adds the reference soundtrack, so every finished MP4 has sound. Silence almost always comes from playback.

Fix:

• Check that the player is not muted, and turn off the silent switch on iPhone.
• Play the downloaded file, not a preview thumbnail in a gallery app.
• If a platform strips or replaces the audio when you upload, add its official "Alors on danse" sound to the post instead.

16:9 looks less like the original

Cause: the reference clip is vertical. 16:9 widens the white studio, and the moves and close-up can follow it less exactly.

Fix: use 9:16, the default, for the closest result and for TikTok, Reels and Shorts.

It's still rendering

Cause: most videos take a few minutes; 30-second videos and busy periods take longer.

Fix: you can leave the page. Your video appears in the history on the generator when it finishes. If it fails, the credits are returned.

Credits were charged but the video failed

Cause: credits are reserved when the video starts.

Fix: nothing to do. If the model fails, a photo is not accepted or the render does not finish, the credits for that video are refunded automatically. Check your balance after the task shows as failed, and contact support if it does not come back.

The photo was rejected

Cause: the file type, size or shape is outside the limits, or the photo shows more than one subject.

Fix: use a JPG, PNG or WebP up to 20 MB with the shorter edge at least 300 px, not extremely tall or wide, and with one person or one pet. Seedance 2.5 also reviews both photos before it renders; if it rejects one, try a different everyday photo.

The watermark is still there after I paid

Cause: the page may still have the old copy loaded.

Fix: reload the generator page. After any purchase, the same download serves the clean original for every video in your history. See how the watermark works.