The portrait is the one source shared by both studio workflows. Before comparing model options, make sure it clearly describes the person you intend to put on screen. This guide is about preparing and inspecting that source—not a promise of generated likeness or motion.
Read the face before the aesthetic
Open the portrait at a comfortable size and ask what is immediately visible. Can you distinguish the eyes and mouth? Is one person clearly the subject? Does a hand, microphone or heavy shadow cover the features you want the system to interpret?
A striking photo and a useful source photo are not always the same thing. A dramatic silhouette might be the right poster image, while an evenly lit portrait makes a better starting point for testing. Keep both files and choose deliberately.
- Start with one clearly visible face rather than a group or collage.
- Prefer a crop that keeps the forehead, chin and sides of the face in view.
- Avoid adding effects just to make the preview look more finished.
Separate the portrait crop from the video frame
The studio concept uses a vertical performance frame, but that does not mean your source portrait has to be a finished vertical poster. Choose the face first. Text overlays, borders and large empty margins are presentation choices, not facial information.
If you prepare a cropped copy, retain the original. Inspect the cropped copy before selecting it: do not accidentally remove the chin or enlarge a tiny face until the detail disappears. The current preview can show your source, but it does not evaluate a future model’s likeness accuracy.
Check the actual file, not only its name
The current studio accepts JPG/JPEG, PNG and WebP portraits up to 20 MB. A supported extension must agree with the file type. Renaming a file from .heic to .jpg does not convert it.
When a file is rejected, create a real supported-format copy using a local image editor, then select that copy. Avoid putting private portraits through an unrelated online conversion service simply to get a smaller upload.
Permission is part of preparation
A portrait of yourself is a straightforward starting point. For anyone else, establish permission to use that person’s image in the intended context. A publicly posted photo is not a general license to place the person into a synthetic performance.
The same permission question applies separately to the reference video, recorded performance and music. Keep a note of what you may use and how you obtained it. Selecting the studio permission checkbox is a confirmation of your decision, not a rights-clearance service.
End with a local check
Select the portrait in the studio and confirm that the preview is the file you intended. Try the video and audio tabs; the portrait is shared during the current page session. Reloading or leaving the page clears those local inputs.
Generation is not connected in the current rap workflow. Preparing a good source is useful work, but the preview itself is not a generated result or a guarantee of what a future model will produce.