Simple Diffusion 0.6.1 - clearer setup and tester diagnostics

Simple Diffusion 0.6.1 — clearer setup and tester diagnostics

Windows x64 early-access prerelease by Prosper Technologies.

What changed

• Copy diagnostics now explicitly includes app version, Windows version/build, NVIDIA GPU, VRAM, driver, RAM and local engine/model readiness.
• Diagnostics use a fixed field allowlist; private fields are omitted and path-like text is redacted. The report is copied locally and is not automatically submitted.
• Added Copy diagnostics directly to first-run setup, so hardware testers can collect a report before installing the engine.
• Setup prominently shows the combined 9.69 GB engine + model download, additional to the small app installer, before anything starts. Sizes now use decimal GB consistently. Existing verified files are reused.
• Added clearer public download documentation, a network-verification guide and a simpler GitHub issue form.
• Help & diagnostics includes a link for reporting inappropriate AI output; opening it does not automatically upload prompts or images.

Download and setup

Download Simple-Diffusion-Setup-0.6.1.exe (approximately 169 MB). No companion installer files are required. Windows setup installs the app only.

When ready, open setup inside the app, review the sizes and model license, then choose Set up & download. The separately downloaded NVIDIA engine is 2.75 GB and the SDXL model is 6.94 GB: 9.69 GB combined, or approximately 9.86 GB including the installer. Allow at least 20 GB free storage plus image storage. Downloads are checked and support pause/resume. Generation works locally after setup.

Updating from 0.6.0

Close Simple Diffusion before running the new installer. Install to the existing location; an uninstall is not required. Ordinary app updates retain the separately installed engine and model files.

Requirements and verification

• Windows 10/11 x64; compatible NVIDIA GPU with 8 GB+ VRAM, driver 560.94+, and 16 GB+ RAM.
• Tested hardware remains Windows 11 / RTX 4080 16 GB. Physical 8 GB and 12 GB GPU testing remains pending.
• 34 automated tests pass, including diagnostics privacy regression tests. TypeScript and production build pass. Packaged version, included changes and release protection checks pass.
• No independent privacy/security audit or new physical-GPU compatibility claim is made.

Current limits

This is an unsigned early-access installer. Windows may show an unrecognized-publisher/SmartScreen warning. Automatic app updates and live pack sales remain disabled. Basic local generation is free. Application source is privately maintained; GitHub source archives contain the public documentation repository.

Installation guide · Privacy policy · Check network activity · Report a bug

Simple Diffusion 0.6.0 - Small App, Optional Engine

Simple Diffusion 0.6.0 by Prosper Technologies

Download only Simple-Diffusion-Setup-0.6.0.exe (169 MB). Windows setup installs just the app, without fetching the image engine.

When ready to create, open setup inside the app. It shows download sizes before you accept the model license and choose Set up & download. The optional NVIDIA engine is about 2.75 GB; the SDXL model is about 6.9 GB. Downloads are verified and support pause/resume. You can close setup and return later.

The engine is installed once in local app data and retained across normal app updates. Existing model files are verified and reused. No manual ZIP extraction, separate Python or CUDA installation is needed. Allow at least 20 GB free plus image storage.

Verified: 27 automated tests, clean first run without an engine, separate engine installation, app-only Windows installation, and packaged actual GPU image generation, export and cancellation.

Unsigned early-access testing build. Windows may show an unrecognized-publisher warning. Automatic updates and live paid-pack sales remain disabled. Requirements remain Windows 10/11 x64, supported NVIDIA GPU with at least 8 GB VRAM and driver 560.94+, and 16 GB RAM. Tested on Windows 11 / RTX 4080 (16 GB); wider hardware coverage remains pending.

Installer registration update: Simple Diffusion is explicitly registered in Windows Settings > Apps > Installed apps under Prosper Technologies, with version 0.6.0 and working uninstall commands. Added an Uninstall Simple Diffusion shortcut in the Start menu. Normal per-user installation and registration checks passed. Close and reopen Settings after installing if the list was already open.

Simple Diffusion 0.5.1 — Hardened release packaging

Simple Diffusion by Prosper Technologies is a free local SDXL image-generation app for Windows with supported NVIDIA hardware.

Unsigned early-access testing build. Windows may warn about an unrecognized publisher. If you prefer a signed application, wait for a signed release. Automatic update installation and live paid-pack sales are disabled.

Download one installer

Download Simple-Diffusion-Install-0.5.1.exe below and open it. This is the only file you need to download manually. Internet is required during installation: setup automatically downloads and SHA-256 verifies about 2.66 GB of installation data from this official release. No companion-file handling, separate Python or CUDA installation is required.

The SDXL model (~6.9 GB) downloads separately in the app. Allow at least 20 GB free plus image storage. SHA256SUMS-online.txt provides the new setup file's integrity hash, not a publisher signature. The old split assets remain hosted because online setup uses them automatically. The app remains version 0.5.1; this changes installation packaging only.

Requirements: Windows 10/11 x64 target, NVIDIA GPU with at least 8 GB VRAM, driver 560.94+, 16 GB RAM. Verified on Windows 11 / RTX 4080 (16 GB). Physical 8/12 GB GPUs and fresh Windows installations remain unverified. AMD, Intel, CPU-only, macOS and Linux generation are unsupported in this release.

Product website · Installation guide · Privacy and licenses

0.5.1 — 2026-10-30 · release hardening

• Release builds now compact the first-party desktop JavaScript and omit source maps and development files.
• The first-party Python worker ships as interpreter-matched bytecode, with a hash checked at application startup. Readable worker.py is absent from the release.
• Enabled Electron archive integrity validation and ASAR-only loading; disabled Node-mode, Node options injection, Node inspector flags and packaged renderer developer tools.
• Added a fail-closed packaging check for private source/build artifacts, worker integrity and required fuse settings. QA/debug launchers cannot enter the installer.
• Added Prosper Technologies copyright notice while preserving third-party notices and licenses.

Verification: 22 automated test groups pass; compacted app and bytecode worker passed actual GPU generation, JPEG export and cancellation; all five collections/13 styles passed import, previews and restart tests. A separate QA executable enabled only its inspector for automation and was removed before packaging. The sealed production executable launched normally, and a modified ASAR was rejected by its integrity check.

These measures deter casual copying and tampering. JavaScript compaction and Python bytecode are not encryption; determined users can still inspect or reverse engineer software running on their PC. Locally installed recipe packs and public model weights remain accessible. No server secrets are bundled. Previously downloaded versions cannot be recalled.

This release remains unsigned; code signing and broader hardware/security review are pending. Automatic updates and live sales remain disabled. October 31 packaging update: added the single-download online installer described above.

Optional NVIDIA Image Engine v1

Optional NVIDIA engine for Simple Diffusion 0.6.0 and compatible later versions.

The app downloads these files automatically after the user chooses setup. No manual ZIP extraction is needed. This engine is installed once and retained across normal app updates. The SDXL model downloads separately.

PyTorch 2.7.1 / CUDA 12.6, Diffusers 0.35.2 and the pinned Python runtime. Third-party notices and licenses are preserved. Requires supported NVIDIA hardware; this does not add CPU, AMD or Intel generation support.