🔍 Image Upscaler

A dedicated page to upscale images in batches right on your own machine — no account, no credits, no internet needed.

Overview

Image Upscaler screen: model settings, scale factor and the image list
The Image Upscaler screen: pick a model, a scale and an output folder, then drop a whole folder of images in for batch processing.

Image Upscaler is a standalone sidebar page (🔍 icon). It uses the bundled realesrgan-ncnn-vulkan engine to enlarge & sharpen images fully locally (offline).

🆓Completely free — works on every plan (including Basic), no credits, no account required. This page is NOT tier-gated.
ℹ️100% local: NO G-Labs/Google account, NO credits, no internet required. Works on any images (your own or downloaded), not just app-generated ones.
⚠️Do not confuse this with the "Image upscale (2K/4K)" option inside the Flow Image / Video creation pages and the plan comparison table — that one is the Flow online upscale and needs a Google ULTRA account + credits. This Image Upscaler page is 100% local & free.

Upscale Models (7 models)

A model-count badge shows in the top-right when the engine is ready. Each model is tuned for a different kind of image:

ModelBest for
StandardBalanced — most photos
Lite & fastFaster, less detail
Digital artDrawings, anime, CGI
High fidelityKeep fine detail
Realistic & sharp (Remacri)Sharp, rich textures
Balanced (Ultramix)Balanced sharpness
Ultra sharpMaximum sharpness
💡All models are 4×-base models for high-resolution output. You still pick an output scale of 2×–8× in the Scale box; 4×/8× outputs can be very large and the engine still produces them.

Configuration

SettingDetails
AI ModelPick one of the 7 models above
Scale2× / 3× / 4× / 5× / 6× / 7× / 8×
FormatSame as source / PNG / JPG / WEBP
Output folderLeave empty = save to the same folder as the source
AdvancedProcessing tile size (Auto), "Better quality (slower)", GPU id
ℹ️If your machine has no Vulkan-capable GPU, the app reports "No Vulkan GPU" — the engine needs Vulkan to run.

How to Use

  1. Load images: drag-and-drop images or a whole folder into the table, or use ➕ Add images / 📁 Add folder
  2. Choose settings: model + scale + format + output folder
  3. Click 🚀 Upscale — batch processing; each row has 5 statuses: Waiting / Processing / Done / Error / Stopped
  4. Review: Compare before/after with a slider (drag the divider, scroll to zoom), Open the output folder, or Retry individual rows
💡Since it runs locally, speed depends on your GPU. Large images at high scale (8×) take more time and RAM.