The quality of your reference photos directly affects your output. The AI reads visual cues from your uploads — lighting, texture, shape, color — so giving it clear, unambiguous images makes a measurable difference.
Product photos
Model photos
Outfit photos
Pose photos
What works well
- Clean background — white, light gray, or transparent (PNG with no background)
- Even, diffused lighting with no harsh shadows across the product
- Sharp focus across the entire product, not just one surface
- Full product visible — don’t crop edges or let the product run off-frame
- Shoot straight-on or at a slight angle; avoid extreme fisheye or wide-angle distortion
What to avoid
- Cluttered backgrounds with other objects
- Heavy shadows that obscure product shape or texture
- Blurry or out-of-focus areas on the product
- Heavy color grading or filters that shift the product’s actual color
What works well
- Face clearly visible and unobstructed
- Even, natural lighting — avoid harsh single-source shadows across the face
- Neutral or simple background so the model isn’t competing with the scene
- Minimal heavy filters or skin-smoothing edits that erase facial features
What to avoid
- Sunglasses, hats, or objects covering the face
- Dark or backlit conditions where facial features are lost
- Extreme angles that obscure half the face
- Heavy makeup filters that remove skin texture
What works well
- Flat lay on a clean, uncluttered surface
- Or: garment on a mannequin, showing the full piece
- Full garment visible — don’t crop sleeves, hem, or collar
- Minimal wrinkles so the AI reads the intended silhouette
- Neutral background that contrasts with the garment color
What to avoid
- Heavily wrinkled or bunched fabric that distorts the garment shape
- Partial shots that cut off part of the garment
- Clothing mixed with accessories or other items in the same frame
- Strong patterned backgrounds that bleed into the garment
What works well
- Full or half body visible depending on the pose you want to reference
- Clear body position — limbs, hands, and head placement all readable
- Good lighting so the body silhouette is distinct from the background
- Neutral background helps the AI read the pose outline cleanly
What to avoid
- Poses where arms or legs overlap ambiguously
- Low contrast between the subject and background
- Motion blur or soft focus on the body
- Cropped poses where key joints or limbs are cut off
General tips for all photo types
- Resolution: Aim for at least 1000px on the shortest side. Higher resolution gives the AI more detail to work with.
- File format: JPEG and PNG both work. Use PNG when your product has a transparent background.
- Color accuracy: Don’t apply heavy filters before uploading — the AI uses your photo’s colors as a reference.
- One subject per photo: Each slot expects one primary subject. Don’t upload collages or mood boards as reference images.
GESTEL processes your reference images through Gemini before generation. Clear, well-lit photos give the model more accurate information to work from.
