Two photos. A 10-second painted duet. Choose who stands on each side.Upload two photos for a painted duet. Your first person faces the camera on the left; your second sings toward them from the right.
Two familiar faces, a geometric painted wall and a duet told through glances.
The idea draws on Gotye's Somebody That I Used to Know, featuring Kimbra: two performers share a geometric painted setting. AI remakes recast the pair while trying to keep the expressions, glances and painted composition. Gotye AI video and painted duet AI refer to this visual remake, not to generating a new song.
Source of the recreated motion referenceThis template uses the displayed 10-second, clothed recreation, not an excerpt of the original music video. It is separate from our generated How to result. Photos guide identity and hair; the reference guides painted staging and costumes. Photo clothing and accessories can still carry through.
The original artwork and the newer photo-remake format are related, but they are not the same video.
Gotye's official video, directed by Natasha Pincus, pairs geometric scenery with body art by Emma Hack. Its still framing and contrasting positions give the duet a distinctive visual identity.
Watch Gotye's official videoA Reddit remake featuring Sam Altman and Dario Amodei prompted discussion about more than resemblance: viewers also noticed whether the background and painted details matched. Those details matter when evaluating a recast duet.
Read the remake discussionChoose a separate photo for the front-facing left role and the right profile role. Start with the fixed short excerpt and review both people through the turn and the side view.
Upload your duo, check the settings and review the finished clip.
Add one clear single-person photo to each slot. Photo 1 is on the left, facing the camera. Photo 2 is on the right, turned toward the first person.
Watch the 10-second reference, choose 480P or 720P, and check the credit total. The model uses the painted scene and costumes as a guide, but clothing and accessories from your photos can appear in the result.
Review the front view, profile, mouth movement and painted wall before downloading the MP4. AI results can vary even when the photos and roles are clear.
Use a clean photo that makes each face easy to distinguish. You do not need to paint your face or recreate the set.
The reference supplies the performance, so you can focus on casting the pair.
Each upload has a matching role image. The prompt assigns photo 1 to the front-facing left performer and photo 2 to the profile performer on the right.
Your images guide each face, head shape and hair. The reference guides costumes and painted styling, but clothes, accessories and other details in the input photos may also carry into the result.
The wall, geometric color patches and framing are part of the reference. The model is asked to preserve them, but painted areas can be partial and their edges may drift. Check both people, especially the side-facing role.
The final clip uses audio from the reference excerpt. Photos do not clone your voice or create a new vocal performance.
This first version is a fixed 10-second clip, not the full music video.
The 480P option costs 150 credits; 720P costs 250 credits. Both use the same two-role, 10-second reference.
Check the generate button for the current total and your available balance. Buying credits does not start a video; a new attempt is a separate task.
Credits are the site's billing unit, not a fixed dollar amount. Higher resolution does not guarantee more accurate likeness, paint or lip movement. Reference-to-video model details
The setup differs across these workflows. Features and prices depend on the individual template or model.
Try another two-person template with its own role mapping and reference.
It is a photo-remake idea inspired by the painted duet in Gotye and Kimbra's music video. This tool uses two photos with a short, clothed recreation of that setting. It does not make a new song or reproduce the complete original music video.
Photo 1 replaces the performer on the left, facing the camera. Photo 2 replaces the performer on the right, turned toward the first. The role images next to the uploads show the positions.
This version requires two separate single-person photos. It does not offer automatic group-photo splitting or a one-role-only edit. Use an unobstructed face for each person.
The photos guide identity, head shape and hair, while the reference guides the clothed costumes and geometric painted look. Neither outfit source is guaranteed: photo clothing or accessories can carry through, and painted details can be incomplete. Review the actual How to example before generating.
The output is a 10-second landscape 4:3 MP4. The price is 150 credits at 480P or 250 credits at 720P. Check the selected total on the generate button before starting.
The shortened name, Gotye AI video and painted duet AI can refer to this same remake format. Different tools use different scenes and lengths. Here, the displayed 10-second reference defines the version you are making.
No. This version generates the video directly from the two photos and the fixed motion reference. It does not include a separate still-image approval step. The How to section shows the exact input photos and an actual reviewed 10-second output, including its limitations.
No guarantee. The side-facing person can lose resemblance, paint can be incomplete or drift, and facial expressions or mouth movement may differ from the reference. Headrests, equipment and marks visible in your photos can carry into the video. Choose clean images and review both people through the whole clip before sharing.
Failed tasks return their generation credits through the site's task handling. A completed clip with visual imperfections is not automatically classified as failed. Check History for task status and contact support if progress appears stuck.
This is an independent AI template, not an official Gotye or Kimbra product. Use photos you can edit and follow the applicable rules for the reference footage and music. Generating a clip does not grant new footage or music rights; disclose AI editing when sharing.
