<aside> 🛠

Auto-Upscale Small Images to the Next Standard Resolution Tier

Bring small or inconsistently sized images up to a standard resolution tier (512, 1024, 1536, 2048, 2560, 3072, 4096) but only when they actually need it. Already big enough? Image passes through untouched. Perfect for normalizing inputs before AI models that expect a minimum resolution, without ever downscaling or duplicating work.

</aside>

What it does

Image Conditional Resize automatically upscales small images to the next standard resolution tier you've allowed. It looks at the input image's width and height, finds the smallest tier that fits both dimensions, and resizes up to it. If the image is already at or above your selected maximum tier, it's returned unchanged. Aspect ratio is preserved by default. Runs instantly with no GPU.

Problem it solves

• Normalize batch inputs – Bring mixed-resolution image sets up to a consistent minimum tier before downstream processing
• Meet model minimums – Many AI models expect at least 512 / 1024 / 2048 px on the short side; this tool guarantees it without breaking aspect ratio
• Avoid pointless upscales – Images already at or above the cap are passed through unchanged
• Predictable tier sizes – Standard tiers (512, 1024, 1536, 2048, 2560, 3072, 4096) align with common model and pipeline expectations
• Cap the maximum size – Set an upper resolution limit so the tool never produces an image larger than you want
• Preserve aspect ratio – Optional aspect-ratio preservation keeps proportions correct after upscaling
• No GPU needed – Pure CPU resampling, suitable for any environment and large batches

Input/Output

<aside>

• Input: Any image
  • Format: PNG, JPEG, BMP, or standard image formats
  • Resolution: Any size (the tool decides whether to upscale based on dimensions)
• Output: Conditionally resized image
  • If both width and height are below a tier: image is upscaled to that tier
  • If image is already at or above the maximum tier: image is returned unchanged
  • Aspect ratio: Preserved by default (configurable)
  • Format: Same as input </aside>

Features

• Resize Method:

  • Bicubic, Bilinear, Lanczos

  Method	Neighborhood	Quality	Speed	Best For
  Bicubic	4×4 pixels	Better	Medium	Balanced quality/speed
  Bilinear	2×2 pixels	Good	Very fast	Quick processing, speed priority
  Lanczos	8×8 pixels (or more)	Excellent	Slower	Best quality, final output

• Maximum resolution tier to scale up to: Caps how high the tool is allowed to upscale.

  • Available tiers: 512, 1024, 1536, 2048, 2560, 3072, 4096
  • Behavior: The tool will only upscale up to (and not beyond) this tier.
  • Examples:
    • Cap = 2560 → an 800×600 image becomes 1024 (next tier that fits); a 2600×2000 image is returned unchanged
    • Cap = 1024 → a 300×300 image becomes 512; a 1500×900 image is returned unchanged (already above the cap)

How to Use

<aside>

1. Load image → Add an Image Loader (or any upstream image source) and connect it to Image Conditional Resize

2. Pick a resize method → Choose Bicubic (default), Bilinear, or Lanczos depending on quality vs speed needs

3. Set the maximum tier → Choose the highest tier you want the tool to upscale to (512 / 1024 / 1536 / 2048 / 2560 / 3072 / 4096)

4. Run Flow → Small images are upscaled to the next standard tier that fits; already-big-enough images pass through untouched

   

</aside>