Photo tips
Bad Times AI Group Photo Guide: Make People Dance Together
Prepare one group photo for Bad Times AI, fit people into a vertical video and check missing subjects, changed faces and overlapping limbs.
Written and maintained by Bad Times AI · Updated
1. Start with one group photo
Upload one JPG, PNG or WebP up to 10 MB containing the people you want in the video. You do not need separate photos for each role. The photo supplies the cast; the template supplies movement. The request asks for a shared scene with the photographed people together.
Use a photo you have permission to process, including permission for the people pictured. The website does not publish a guaranteed maximum number of performers.
2. Check every face before uploading
Zoom in on the original. Each face should be distinguishable, rather than blurred, hidden behind another head or reduced to a few pixels. Prefer even lighting, visible body outlines and space between arms. Avoid tight hugs and overlapping hands when you want motion to remain easy to follow.
A wide row of people must fit into a 9:16 output. This can make faces smaller or require the scene to be rearranged. A compact group with room around the edges gives a simpler composition to work with; no framing choice guarantees that everyone will be preserved.
3. Choose the dance and confirm the order
Watch the full reference rather than matching its cast count to your photo. A duo or split-screen reference does not require two uploads. Select the movement you want, upload your group photo and check the preview before payment.
Guest checkout needs no account. Review the current pack and total before submitting, then return on the same device and browser. Keep the private result link so you can check the existing order instead of buying again if processing takes longer than expected.
4. Review each person through the whole video
Count the visible people at the beginning, middle and end. During turns, check for swapped faces, duplicated subjects, merged bodies, missing people and limbs belonging to the wrong person. The request asks the model to preserve identities, but that instruction is not a guarantee.
The historical duo output below was generated with Seedance. It is a completed two-person example, not proof that every group size, source photo or current-model attempt will behave the same way.
If someone is missing or the group is too crowded
For a future attempt, choose a less crowded image with larger faces and fewer obstructions. Re-cropping must not remove the people you want to include. If the original is too blurry to identify a face, choose a better photograph instead of enlarging that blur.
Check the order status before retrying. Confirmed failures restore the applicable deducted allowance; dissatisfaction with a completed result does not automatically trigger that classification or a cash refund. A new generation is a separate request. Save the MP4 locally after reviewing it with sound.
Completed duo example · Seedance
These are existing completed generations, not template reference previews. The original input photos are not displayed here, so this is not a before-and-after comparison or a promise of identical results.