v0.0.5

Changed

• Enhanced UI responsiveness by refining server parameter controls in sd_gui.js, improving the locking behavior of model and server settings during active inference [1]. Parameters such as model type, sampler, and threads are now frozen during generation and auto-thaw on unload, ensuring stable server state without user intervention.

Fixed

• Resolved UI display inconsistencies and status indicators, ensuring accurate progression from “Offline” → “Starting...” → “Online” → “Busy” (while generating) → “Online” during server lifecycle.
• Corrected layout and rendering issues in app.py and css/styles.css to improve visual consistency across interfaces.

Added

• Added .gitignore entries to exclude 'models/' and 'output/', improving repository hygiene.

This release delivers a more reliable and intuitive server-mode experience, with critical parameters locked during active use to prevent unintended changes, while keeping per-generation inputs (prompt, steps, CFG scale) fully editable. The changes are non-intrusive, with zero side effects in core logic.

SD.cpp-WebUI v0.0.3.5

🚀 Server Mode UX

Key Features:

• ✅ Model dropdowns FREEZE when sd-server starts
• ✅ Server params (seed/sampler/etc.) lock during runtime
• ✅ Prompts + sizes stay LIVE for per-generation tweaks
• ✅ Loading state FREEZES (Unload button for cancel)
• ✅ Auto-thaws on Unload

Minimal changes, no known side effects.

v0.0.3.1

This release consolidates image-to-image generation into the txt2img tab, eliminating the standalone img2img interface for a cleaner user experience.
Key features include:

• Support for uploading and automatically converting any image format to PNG with transparency via the /upload_init_image.
• Integration of init_img and strength parameters to control denoising levels during image generation
• Automatic random seed generation when seed is set to -1 using the random module
• Addition of Flux2 model configuration in config.json, expanding model options
• Improved system stability with updated docker-compose.yml, including a /dev/urandom mount for entropy
• Dependency on Pillow added to requirements.txt to enable robust image processing
• Refactored JavaScript and HTML to remove redundant UI elements and streamline interactions
• Enhanced logging with payload truncation to prevent log bloat from large base64 image data

This release significantly improves usability, performance, and maintainability of the SD.cpp-WebUI interface.

v0.0.3-alpha

Release Note – v0.0.3‑alpha
Release Date: 2026‑01‑15

---

🎉 Highlights

Feature	Status	Notes
Server Mode	Largely complete	Supports Flux, Z_image, and SD3.5 models.
CLI Mode	Fully functional	Works with Flux, Z_image, and SD3.5.
Video Mode (Wan)	Experimental	Video generation is available but still in early testing.
Upscale Mode	Planned	Not yet released.
Additional Image Models	Planned	Future releases will expand the model library.
img2img	Forthcoming	Image‑to‑image functionality is under development and will be released in a later version.

---

⚠️ Known Issues

• Random Seed Handling – In server mode, the seed parameter does not reliably produce different results for each request. Setting the seed to -1 does not trigger the expected randomization behavior. This inconsistency is documented and will be fixed in an upcoming patch.

---

🔧 Future Work

• Server Mode Improvements – Implement robust seed handling to guarantee deterministic generation.
• Upscale Mode – Develop an image upscaling pipeline for higher‑resolution outputs.
• Expanded Model Support – Add more high‑performance image generation models to the library.
• img2img – Complete the image‑to‑image feature and integrate it into the CLI and server APIs.

---

📦 Installation

# Using pip
git clone https://github.com/taltoris/SD.cpp-WebUI
git clone --recursive https://github.com/leejet/stable-diffusion.cpp
# follow build instructions for stable-diffusion.cpp
# more info in README.md

---

Enjoy! If you encounter any problems or have feature requests, please open an issue on our GitHub repository.