SKUVis

中文说明筹备中 · English content below

IMAGE TOOL

Smart image pipeline in one run

When your cleanup recipe is fixed, run detection through upscale in one quote instead of opening six separate tools.

结论：Select one or more vision abilities for a single product photo; SKUVis runs the Aidge VisionFlow pipeline and returns one processed image after an 8-credit quoted run—inputs must have each edge greater than 256 px, long edge ≤1920 px, short edge ≤1080 px, and file size ≤5 MB.

Smart images abilities around one product photo
One upload, one combined quote—final image only (no per-step downloads).

Abilities you can combine

Toggle any subset of six steps: find elements, cut out, erase, translate, crop, upscale—each selected ability must have its sub-options filled.

Six smart image abilities and vision-flow limits
Example: find + cut out + crop for a square white main from a messy supplier shot.

Abilities must be unique in one run. Translation requires an allowed language pair; erase requires elements or a mask when ability 3 is on.

  • 1 Find elements — watermark, logo, text, text block (on product vs background)
  • 2 Cut out — white or transparent matting, optional keep-matching filter
  • 3 Erase — element types or painted mask
  • 4 Translate — source/target language pair supported by smart-images list
  • 5 Crop — default 800×800 target (editable 100–5000)
  • 6 Upscale — factor 2–4

When to use smart images vs single tools

Use smart images for repeatable combo jobs under vision-flow limits; use standalone tools when you need to approve each step separately.

Limits are tighter than background removal (3000 px) because VisionFlow enforces the 1920×1080-class cap and 5 MB max.

How it works in SKUVis

Upload within vision-flow limits, configure abilities, confirm 8 credits, download one output from History.

  1. Upload a photo passing validateVisionFlowImage checks.
  2. Select abilities and fill conditional fields (detect types, matting mode, erase mask, languages, crop size, upscale factor).
  3. Review the 8-credit quote and submit.
  4. Inspect the single output before listing—retry with standalone tools if a stage needs manual QA.

There is no partial credit if you dislike one stage; split into separate tools for step-by-step control.

What it is not

Smart images is not product-images planning, virtual try-on, or video.

Failed pipeline runs follow the same credit release rules as other tools when applicable.

Inputs & outputs

What you provide

  • One product image
  • One or more abilities (1–6) with required sub-fields

What you get

  • One processed image
  • Each edge >256 px
  • Long edge ≤1920 px, short edge ≤1080 px
  • 5 MB max
  • 8 credits on current preset—confirm in app

Common use cases

  • Detect, erase promo text, mat white, crop square in one launch-day pass
  • Translate and crop when dimensions fit vision-flow caps

FAQ

Why are limits tighter than background removal?

VisionFlow shares the 1920×1080-class and 5 MB constraints enforced before the provider call.

Can I see intermediate files?

No—only the final image. Use single-purpose tools for step-by-step previews.

If the pipeline fails, are credits returned?

Failed or canceled runs show as not charged in History when release rules apply.