1. Assign identities to performers, not screen positions
Open the APT studio and prepare one portrait per performer. The visible demonstration labels Image 1 as the performer in the gray jacket and Image 2 as the performer in the black shirt. The slot labels describe the displayed example identities so you can choose who takes each role.
The underlying performance changes shots and positions. A person who begins on one side can appear somewhere else later. Check the whole demonstration, including singing, dancing and drumming, rather than treating the upload order as a permanent left/right split.
2. Upload one clear face per image
Use separate JPEG, PNG or WebP images up to 10 MB each. Give each photo one obvious subject with the full face visible. Uploading the same group image to both slots makes it harder to tell which identity belongs to which part.
Compare the selected thumbnails with the role labels before generation. Use Swap portraits if you want to reverse the casting. Confirm permission for both people, including parent or guardian authorization when a person is under 18.
3. Choose the performance length and crop
The 5, 12 and 26-second choices all start at 0:00. A longer cut includes more of the existing performance, not a separately generated introduction. Choose 480p or 720p and inspect the exact quote for that source length.
Original framing retains the landscape scene. A centered 1:1 or 9:16 crop can cut off one performer, especially after the identities change position. Watch the crop outline across the relevant movements before selecting a narrow output frame.
4. Keep the AI demonstration and completed case in context
The public studio demonstration shows example AI identities. Your result uses your own portraits and selected frame. The lower completed case is a separate finished example; its faces and public poster do not become the private input for your generation.
For APT, the official footage remains a private generation input. Watching the AI demonstration helps explain the scene and role mapping without exposing that source in the browser.
5. Review singing close-ups and the second performer
After reviewing and confirming the request, wait for the private result page to report completion. Check each identity separately through movement, changes of angle and mouth close-ups, then watch the two people together.
Listen to the restored source audio and inspect mouth timing. Download the MP4 after the video is ready. If you try another portrait or repeat a completed generation, treat it as a new request with its own credit quote.
Reviewed against the EasyVideoAI studio on October 19, 2026. Check the live studio for current availability and the final credit quote.