Skip to main content

Add a Costume Reference Image

Goal: Upload a photo of the exact costume look you want so the AI has a visual anchor — not just a text description — when generating the character's appearance.

A text description tells the AI what to do. A reference image shows it. When you supply both, the AI uses the image as a visual constraint that keeps renders consistent across shots and scenes — critical for a character who appears in dozens of shots across a film. Reference images are especially useful for unusual fabrics, distinctive patterns, non-standard silhouettes, or any costume element that is hard to describe in words alone.

Steps​

  1. Open the character's Costume tab. In the Casting panel, click the character, then click the Costume tab. If a default costume is already defined, you will see the current description. If not, follow Define a Character's Costume first.

  2. Click Add Reference Image. The image upload dialog opens. You can drag and drop a file or click Browse to select from your file system.

  3. Upload the file. ACT3 AI accepts JPG and PNG files. For best results, use a photo with clear lighting, a neutral or solid-color background, and a full or three-quarter view of the costume. Avoid motion blur, heavy shadows across the garment, or backgrounds that compete with the clothing.

  4. Tag what the reference covers. After the upload, a tag selector appears. Choose one or more scope tags: Full Outfit, Top Only, Bottom Only, Shoes, Accessories, Headwear. These tags tell the AI which parts of the costume this image should constrain. A reference tagged Shoes will affect footwear in shots but will not override how the AI interprets the jacket.

  5. Add a text note explaining what to match. Click into the Match Note field and describe what is most important to preserve from this image. Example: "match the jacket color and lapel width exactly — the double-breasted silhouette is intentional". Keep it to one or two sentences.

  6. Preview how the reference appears in the AI prompt. Click Preview Prompt to see how the reference image and its match note are bundled into the shot generation prompt. This shows you exactly how the AI will receive the instruction.

  7. Add multiple reference images for different costume elements. If your character's costume has distinct pieces that are each visually important, upload a separate reference for each. Common splits: one image for the full outfit, one close-up for the shoes, one for an accessory (watch, bag, jewelry). Each image gets its own scope tag and match note.

  8. Remove a reference image. To delete a reference, click the three-dot menu on the image thumbnail and select Remove Reference. The costume text description remains intact. Remove images if they are outdated (for example, after a costume change) or if you are getting inconsistent renders because two references are conflicting.

Tips​

  • Use real wardrobe photos or fashion reference images, not AI-generated images. AI-generated references introduce compounding distortion — the generation model interprets an already-interpreted image, compounding uncertainty in small details like buttons, fabric texture, and fit.
  • A photo taken against a plain wall or photography backdrop is worth more than a stylized editorial shot with dramatic lighting. The AI needs to see the garment clearly, not the art direction.
  • If you cannot photograph the actual costume, a product photo from the brand's website (or a mood-board image from a fashion reference site) works well as a stand-in.

Next steps​