Download

ImgSera for Windows

Two ways to run ImgSera on a 64-bit Windows machine. Both run entirely locally — installing the application does not upload or synchronise anything.

Available releases

No release files are published yet

The official ImgSera downloads are still being prepared. The controls below are deliberately disabled rather than linking somewhere that does not work. This page updates as soon as the files are published.

Windows Installer

Coming soon

Standard setup package. Installs ImgSera for the current user with Start Menu and desktop shortcuts, and provides a real uninstaller.

File
.exe
Version
1.0.0

Not published yet. This control activates when the official windows installer is released.

Portable Version

Coming soon

A single self-contained folder you can run without installing. Ideal for USB drives, locked-down machines and multiple side-by-side versions.

File
.zip
Version
1.0.0

Not published yet. This control activates when the official portable version is released.

Version numbers are the current release series

File sizes, checksums and release dates are published only once a build is actually released. Track every documented change on the changelog page: /changelog.

AI models

ImgSera's AI tools run on real ONNX models that you download once, on request. They are listed here by friendly name; everything else — size, version, licence and checksum — comes straight from the application's own model manifest.

ClearRealityV1 — Fastest

Upscaling

A 1.6 MB SPAN ×4 upscaler that runs on any CPU. No GPU required.

  • Size 1.7 MB
  • Tier Very low
  • Version 1.0.0
  • Licence Apache-2.0 ✓ verified
  • Runs on CPU only
  • Offline Yes, after download
  • Smallest download — runs on any CPU, no GPU needed
Integrity & source
SHA-256
4b03264eb2f3b1b5a5b253c850045a94bbc3e4aac3946055561e120f1adf0e58
Download

1.7 MB package

U²-NetP — Fastest

Background removal

Very low resource usage — the fastest option, for quick simple removals.

  • Size 4.6 MB
  • Tier Very low (default)
  • Version 1.0.0
  • Licence Apache-2.0 ✓ verified
  • Offline Yes, after download
  • Very low resource usage — fastest processing
Integrity & source
SHA-256
17a3a4c26ae2bfbfab7a3936f3496200a7e91b680f9090b29d70c7a593bd586d
Download

4.6 MB package

HAT-S — Advanced

Upscaling

Hybrid-attention transformer super-resolution at ×4, for maximum detail.

  • Size 21.8 MB
  • Tier High
  • Version 1.0.0
  • Licence CC-BY-4.0 ✓ verified
  • Offline Yes, after download
Integrity & source
SHA-256
6c7097c29cb92fe4fb7ed8358c4cf35e018aab2ca63691444f5a613ceb346cde

Not published yet · 21.8 MB

SwinIR — Balanced

Upscaling

Transformer super-resolution at ×4, with strong texture fidelity.

  • Size 48.3 MB
  • Tier Medium
  • Version 1.0.0
  • Licence Apache-2.0 ✓ verified
  • Offline Yes, after download
Integrity & source
SHA-256
3590eb306b1c5aa6c03aac2fb2a58f24c087f95894d5cfbabf8f9f0405e5569c
Download

48.3 MB package

Real-ESRGAN Upscaler

Upscaling

The Real-ESRGAN ×4 restoration network. Enlarges images at 4×, with 2× available through high-quality resampling.

  • Size 67.3 MB
  • Tier Standard
  • Version 1.0.0
  • Licence BSD-3-Clause ✓ verified
  • Offline Yes, after download
  • 4× AI upscaling — 2× via high-quality resampling
Integrity & source
SHA-256
58afe38d33dc1056e1b412fc2c304e52eb8158be2b742a514378826f0583d9c3
Download

67.3 MB package

U²-Net — General

Background removal

U²-Net salient-object segmentation for one-click background removal, with better detail than U²-NetP.

  • Size 181.9 MB
  • Tier Low
  • Version 1.0.0
  • Licence Apache-2.0 ✓ verified
  • Offline Yes, after download
  • Low resource usage — fast, general-purpose processing
Integrity & source
SHA-256
a663b159b960c8010bdfe0f9ab4bff99582d1c3274739a3d08e31e645dc950f9
Download

181.9 MB package

BiRefNet-HR — High Resolution

Background removal

High-resolution background removal for edges, hair, fur and complex foregrounds.

  • Size 701.3 MB
  • Tier High
  • Version 1.0.0
  • Licence MIT ✓ verified
  • Offline Yes, after download
  • Dedicated GPU recommended — 8 GB VRAM minimum, 12 GB+ for best performance
  • AI working resolution up to 2048 px; output keeps the original image size
Integrity & source
SHA-256
4b6b73e91ab688c8f48004b6a96cdd8a349a4f0dc2099e75c4068004f152d8e2

Not published yet · 701.3 MB

BiRefNet — High Quality

Background removal

High-quality bilateral-reference segmentation — better on complex edges, at a higher resource cost.

  • Size 890.5 MB
  • Tier Medium
  • Version 1.0.0
  • Licence MIT ✓ verified
  • Offline Yes, after download
  • Dedicated GPU recommended — 4 GB VRAM minimum, 6 GB+ for best performance
  • AI working resolution up to 1536 px; output keeps the original image size
Integrity & source
SHA-256
45c2787a17692126861b7de32291fb13e09ca4c8ac4450368278362f7fe4d5aa

Not published yet · 890.5 MB

How models are installed

Models are not bundled with the installer. In the application, open AI Models and download the ones you want — or accept the one-time prompt when you first use an AI tool. After that they work offline, and the application tells you which execution provider is active.

System requirements

What ImgSera needs depends on what you are doing. Requirements marked provisional have not yet been measured on the stated hardware.

Basic image editing

Provisional

Converting, resizing, compressing, PDF tools and batch processing.

Operating system
Windows 10 or 11, 64-bit
Processor
Modern dual-core CPU or better provisional
Memory
8 GB RAM provisional
Storage
About 2 GB free provisional
Graphics
No dedicated GPU needed — editing, compression and PDF work are CPU-bound

The operating system and the "no GPU needed for core editing" statement are confirmed by the application itself. The processor, memory and storage figures are provisional and have not yet been measured on the stated hardware.

Large images and batch processing

Provisional

Big files, high-resolution output and long batch runs.

Operating system
Windows 11, 64-bit provisional
Processor
Modern six-core CPU or better provisional
Memory
16 GB RAM or more provisional
Storage
SSD with about 10 GB or more free working space provisional
Graphics
A dedicated GPU helps, but is not required for these tools

Requirements rise with image resolution, image count, output dimensions and how many images run at once. Every figure on this tier is provisional — we have not yet benchmarked large-image or batch workloads on this specification, so treat it as a target rather than a tested guarantee.

AI features and models

Provisional

Upscaling, denoising, enhancement and background removal.

Execution provider
DirectML on a DirectX 12 GPU, with automatic CPU fallback
GPU requirement
Not required — every AI feature also runs on the CPU, more slowly
Memory
Depends on the model and input resolution — see each model below provisional
VRAM
Only the two BiRefNet models state a figure: 4 GB / 8 GB minimum
Extra storage
Each model is downloaded on demand; sizes are listed per model below
Offline use
Yes — once downloaded, models run with the network disconnected

Requirements depend on the model chosen, the input resolution, which execution provider is active and how much memory is free. Not every model states a VRAM figure, and none of the published models requires an NVIDIA GPU specifically.

What happens after you install

ImgSera opens straight into the dashboard. There is no sign-up step and no configuration wall.

  1. Import your images

    Drag and drop files onto the window, use Add Images, import a folder including subfolders, or paste from the clipboard. Every file is validated by content before it is decoded.

  2. Pick a tool

    Choose a task such as Convert, Resize, Compress or AI Enhance. With a selection active the tool processes that selection; with nothing selected it processes everything you imported.

  3. Export

    Choose an output folder and start the job. Files are written atomically, so a cancelled batch never leaves a corrupt export behind.

AI models download on request

Upscaling and background removal need an AI model the first time you use them. ImgSera tells you when one is missing and offers to download it. After that, those tools work with the network disconnected.

Not published yet?

This page is the single source of truth for ImgSera releases. Each entry updates here as soon as the file goes live.

Releases are being finalised. The download page shows the current status honestly.