API limits per generation request: 60 MB per file, 200 MB total per request. Per-template slot limits add additional constraints. If you hit a 413 or a client-side cap, here are the limits and how to fix it.
API limits
- Per file: ≤ 60 MB.
- Per request total: ≤ 200 MB across all attached files.
Per-template slot limits
- Ghost mannequin / image-creation (non-batch): 3 (Flash) or 10 (Pro / nano-banano-2).
- Virtual try-on: exactly 2 images.
- Styled photoset: ≥ 2, ≤ 10 (first = creator frame, rest = garments).
- Reference photoshoot: ≤ 10.
- Batch (image-creation, ghost-mannequin, ai-photoshoots): up to 50.
- Edit: 2–7.
- Edit-by-prompt: 1–2.
- AI model: exactly 1.
Some UIs are stricter than the API
A few client UIs apply a 50 MB per-file cap before upload (eg. Batch Creation, Cabinet, Post Production Filters), even though the API accepts up to 60 MB. Compress before upload to stay safe across all UIs.
Batch Creation staging budget
Batch Creation stages originals in the browser's IndexedDB before upload. The staging budget is 15 MB total (per browser session). If you hit that limit, downsize the originals before adding to the batch — the AI does not need huge originals; 1500 px on the longest side is plenty for catalog inputs.
How to shrink files
- JPEG quality 85, longest side 2000 px — most fashion garment photos compress to under 1 MB at this setting with no visible loss.
- PNG → JPEG for photographic content. PNGs are 5–10× larger.
- On macOS: Preview → Tools → Adjust Size + Export as JPEG. On Windows: Photos app > Resize.