PixelOkayImage tools
Tools

PIXELOKAY / IMAGE STUDIO

AI Image Upscaler

Enlarge photos, screenshots, illustrations and pixel art privately. Precise resizing, sharpening and batch export work without AI; optional local neural enhancement is available on demand.

Private by designYour images stay on this device. Only the free AI model and runtime are downloaded after you explicitly approve Model Setup.

Choose an image

Drop, paste or select a PNG, JPEG, WebP, BMP or GIF. GIF uses its first frame · 24 MB maximum.

Make a small image bigger

Choose a file or try the sample. Resize, enhance, inspect and download without AI, a signup or an upload.

Batch upscale · No AI

Queue images and process sequentially with the current Standard settings. Up to 12 images.

Local processing

No images queued. Individual exports remain available if a batch ZIP exceeds the 128 MiB memory limit.

    How the free image upscaler works

    1. Add an image

    Import a photo or illustration. The file is decoded and processed locally; no image file is uploaded.

    2. Resize without AI or opt in to AI

    Standard mode processes pixels directly in the browser. For optional neural enhancement, explicitly choose AI and approve model loading first.

    3. Inspect and download

    Compare before and after, choose PNG/JPEG/WebP, and save your result. Processing can be canceled safely.

    Using AI Image Upscaler

    Import images, choose Standard (the default), select resizing and enhancement settings, then apply. Batch process several images without AI. Inspect the before/after result before downloading. A supported local model file can replace the model download but does not supply the runtime.

    AI downloads a model/runtime and can be slow or exceed device memory. The neural model produces 4× output; 2×/3× downscale it, and 6×/8× use conventional enlargement afterward. Standard mode performs no neural enhancement. AI textures are inferred rather than recovered evidence. PNG/WebP alpha can show matte halos; JPEG flattens to white. Animation and EXIF/ICC retention are unsupported. Optional local runtime files are not bundled.

    References: Real-ESRGAN ONNX model · Real-ESRGAN project · ONNX Runtime

    Related tools

    LOCAL AI SETUP

    Set up AI on this device

    First, a model needs to be loaded on your device

    This tool uses a free Real-ESRGAN 4× ONNX model on your device. After you click Download & Load Model, the browser downloads model and runtime assets from public hosts. Your image pixels are processed locally, not sent to an inference server.

    1. 1. Check device
    2. 2. Download / cache
    3. 3. Initialize Real-ESRGAN
    4. 4. Upscale image
    Real-ESRGAN General x4v3 is a 4× neural image upscaler, not a chat model. Standard resizing works without AI.

    Device & model resources

    Measurements may be approximate
    Device RAM / model memoryChecking device information…

    Exact live RAM/GPU memory cannot be measured in most browsers. The upscaler uses ONNX Runtime Web on CPU/WASM in a worker.

    Browser website storageChecking storage…

    Site-wide estimate; includes other PixelOkay tools and is not RAM.

    ONE ACTION · ALL RECOGNIZED AI MODEL DOWNLOADS

    Free AI model storage across PixelOkay

    Scan recognized same-origin WebLLM, Transformers.js, ONNX Real-ESRGAN and other verified model caches, including downloads from other PixelOkay tools. Saved website data and browser databases are not cleared.

    Cache status has not been checked yet.

    Run a scan to see how many model caches can be cleared together.

    Browser-native models and unknown IndexedDB/OPFS or private runtime caches may remain. Unknown application/user storage will not be deleted. Other running tools might redownload caches; close their tabs first.

    Having trouble downloading or loading AI?

    If model loading fails, refresh device and storage details, retry, or inspect recognized model caches. Check your connection and WebAssembly support. Standard resizing remains available without any AI.

    Only model caches are targeted. Other PixelOkay tools may need to redownload their models. Never clear all website data without exporting important chats and projects first.

    Select Check compatibility to verify WebAssembly and Worker support.
    Technical details and diagnostics

    Runtime: ONNX Runtime Web 1.24.3 · Real-ESRGAN x4v3 · WebAssembly / CPU · pinned SHA-256 verification. Model setup uses a worker; image processing is local.

    CROSS-TOOL MODEL CACHE CLEANUP

    Delete recognized AI model downloads?

    Recognized model cache targets will be listed here.

    Only independently recognizable model caches are targeted. Other PixelOkay tools may need to redownload models. Saved image settings, source image, files, and browser databases are not selected, but unknown stores are not inspected or removed.

    Close other AI tabs before proceeding. Browser-managed, IndexedDB, and OPFS model files may remain.