The source image does more than decide what the pet looks like. It guides the silhouette, colors, clothing, accessories, and the details that need to remain recognizable across every motion state. A simple image gives the model room to be precise.
Start with one obvious subject
Choose an image where the person, pet, logo, or object is immediately easy to identify. Crowded group photos and busy scenes force the model to guess which details matter.
- Keep the full head and main body visible
- Avoid heavy foreground objects or hands covering the face
- Crop away unrelated people and background clutter
Front-facing is the safest starting point
Side views can work, but a front or near-front angle gives clearer information about symmetry, facial features, clothing, and proportions. Extreme camera angles often become inconsistent once the character turns, runs, or jumps.
- Use eye-level or slightly elevated photos
- Prefer relaxed poses over dramatic action poses
- Leave a little space around the subject
Good light preserves the details you care about
Soft, even lighting helps the generator separate the subject from the background. Very dark images, colored stage lighting, and strong shadows can change coat colors, skin tones, or small accessories.
- Natural window light usually works well
- Avoid blur and aggressive beauty filters
- Use the highest-resolution original you have
Choose what must stay recognizable
Before uploading, identify two or three details that make the subject unmistakable: a hat, coat color, hairstyle, glasses, or a simple logo shape. Small jewelry and tiny printed text are less reliable across dozens of motion frames.
Use one clear subject, a front or near-front angle, even lighting, and a simple pose. Make the important visual details large enough to survive animation.