v0.2.1

Release Notes

v0.2.1

This version adds an AI prompt enhancer and a built-in image editor. You can now have a local language model rewrite your prompt into something richer with one click, and resize or crop images right before adding them to a session — without ever touching the original file.

Added

• Enhance prompts with AI — a new Enhance button in the chat composer rewrites your current prompt into a richer, more detailed one using a local AI language model. The model and its runtime download automatically on first launch, and you can manage the model catalog, system prompt, and inference settings in Settings (Language Models and Language Inference). While an enhancement runs, the button pulses — click it again to cancel.
• Resize and crop before adding — when you add input or reference images, you can now edit them first in a full-screen editor: crop to any region, zoom, rotate, and pan (middle-click), pick a resolution preset, or set a custom size. Your original file is never modified, and multi-image picks are processed one at a time (Save & Next / Skip).

Improved

• Latest sd-cli runtime — the bundled generation runtime is updated to the newest stable-diffusion.cpp builds (1054–1056).
• LTX-2.5 GGUF models — the Vantage with AI LTX-2.5 GGUF models are now correctly listed as video generation models in the catalog.
• Enhance button feedback — the busy indicator now uses a calmer pulse animation instead of a spinner.

Fixed

• Prompt box sizing — the composer text box now resizes correctly when the prompt is changed by the app (for example, after using Enhance), so the full text is no longer clipped.

v0.2.0

Release Notes

v0.2.0

This version adds support for the newest generation of models and adds powerful new ways to create video. You can now generate short AI videos with the WAN family, produce high-quality video with LTX 2.5, and expand a picture's borders. Video output is now converted to a modern format that plays directly in the app, and the model library is organized more cleanly so files never get mixed up. A new resolution picker makes it faster to pick the right size.

Added

• Generate AI videos — a new video mode (WAN support) creates short video clips from a text prompt or an image. It supports companion "expert" models for higher quality, audio tracks, control-video input, and end-image frames, with adjustable length and speed.
• Play videos in the app — generated videos are converted to a modern MP4 format and played inline with native controls (play, seek, volume), so you no longer have to open an external player.
• LTX 2.5 video — support for the LTX 2.5 video model family, which turns a text prompt or image into a high-quality video clip. It can use a companion audio model for synced sound and a latent upscaler for sharper, higher-resolution output.
• Pick a resolution in a dialog — a new unified resolution picker replaces the small dropdown, making it easier to browse and choose a size.

Improved

• Cleaner model organization — models are now stored in a family-scoped layout with a resolution chain that checks the right file first, so different models can never overwrite or confuse each other.
• Better model naming — models show clear, human-readable names (including MOSS-Image-SFT), and catalog entries are matched more reliably.
• Smaller, faster videos — video output is automatically converted to H.264 MP4 with fast startup, so playback is smooth.

Fixed

• Reliable model resolution — a wrong-family or wrong-source file can no longer be silently loaded in place of the correct model.

v0.1.3

Release Notes

v0.1.3

This version expands the kinds of models you can use and adds powerful new ways to edit images. You can now paint over a photo to change it, mask an area to replace just that part, or gently expand the edges of a picture. New model sources bring more choices, and you can control your images even more precisely by locking in the random seed for the same result every time. It also checks for new models automatically when you open the app.

Added

• Paint directly onto a photo — a new brush tool lets you draw over an image to change or cover parts of it, then send it along with your request.
• Edit a part of an image — you can now mask (highlight) a section of a photo so only that area gets redrawn, or expand the edges of a picture to make it larger. This works for models that support it.
• LoRA support — LoRA files are small add-ons that teach a model a new style, look, or character. You can pick them from a gallery and apply them per chat session, so you don't have to re-choose them each time.
• More model sources — the app now pulls models from CivitAI, a popular site for sharing AI models, in addition to the existing source. This gives you access to a much wider library of models.
• See model sizes at a glance — each model now shows how much space it takes up, so you can tell a small file from a large one without guessing.
• Control the random seed — each image starts from a random number. You can now pin that number (the "seed") so a good result can be repeated later, or roll a fresh one with a single click.
• New model types — support for additional model families such as Flux and Moss has been added, along with a new "edit" mode tuned for changing existing images.
• Tokenizer support — some models need an extra helper file to read your words correctly. The app can now download and use these files automatically.
• Flow-shift setting — a new quality slider (available on supported models) that helps the model produce cleaner, more detailed images.
• Automatic update check — when you open the app, it quietly checks for newer model and runtime settings and gently reminds you when something new is available, so you can keep your options up to date without extra effort.
• Manage private-model access — a new section in Settings lets you add, edit, or remove access tokens for private or paid model sources, so you can download models that need a login. Tokens can be given broad or narrow access with simple wildcards.
• Search tokens — managing access tokens is easier with a dedicated, searchable list in Settings → Tokens.

Improved

• Better model organization — the model browser was reworked so it can support many different model sites cleanly, and models now show clear, human-readable names instead of long file-style codes.
• Clearer error messages — when a generation fails, the app now records the details in the log and shows a simple reminder, instead of dumping confusing technical text into your chat.
• Cross-platform builds — the app can now be built for Linux and macOS as well as Windows, bringing it to more computers.
• New website — a landing page now showcases the app's features, screenshots, and a video demo, making it easier to learn what it can do.
• Smoother downloads — background checks for new settings run quietly at startup without interrupting your work, and failed checks are treated as "nothing new" rather than an error.
• More models to choose from — the built-in catalog grew with new providers and model variants, so there are more styles and sizes to pick from.

v0.1.2

Release Notes

v0.1.2

This version adds several new model families, lets each session remember its own model and resolution, and lets you build images from your last result. It also adds finer control over how images are generated — choose your sampler and scheduler, write a negative prompt, and override sampling per session.

Added

• More model families — the app now ships support for SD3.5, Krea 2, SDXL, and Anima models alongside the existing ones, so you can work across a wider range of diffusion types.
• More models to choose from — the built-in model catalog is expanded, adding new SD3.5, Krea 2, SDXL, and Anima options from additional providers, so there are more quantizations and sizes to pick from.
• Reference image from your last generation — when a model supports image-to-image, you can start a new image from the one you just generated. This is available only for models that opt in, and controls how much the original is changed.
• SD3.5 multi-encoder support — SD3.5-style models that bundle several text encoders are now handled automatically, resolving and passing each encoder file instead of a single one.
• Sampler and scheduler selection — choose from a set of samplers and schedulers in Settings → Sampling instead of relying on the default. The chosen scheduler is only sent to the runtime when it is explicitly set, so models that prefer their own default keep working out of the box.
• Negative prompt — enter a negative prompt in Settings → Sampling, or override it per session from the chat composer. It is only sent when you actually type something.
• Per-session sampling override — each chat session can now keep its own sampling settings (steps, CFG, strength, sampler, scheduler) on top of the model defaults, so you can fine-tune one session without affecting the others. The current override shows in the session summary.
• Remember model and resolution per session — each chat session now keeps the model and resolution you last chose. Switching to a session restores those inputs, so you don't have to re-pick them every time.
• App logging — Settings → Logs shows the latest application log on demand, which helps diagnose problems. Logs are written and rolled automatically so they don't grow without limit.
• Save image-as with a quick open-folder action — after saving a generated image, a toast offers to open the folder containing it.
• Resume downloads — pause, cancel, or retry a download anytime. Interrupted downloads won't leave broken or incomplete files behind.
• Update models and settings from the app — Settings → Runtime and Settings → Models now have Check and Reset buttons. Check downloads the latest available models and settings from the internet, so you don't need to reinstall the app to get new options. Reset restores the original built-in options.
• Better model selection — a new searchable list replaces the old dropdown, showing a short description for each model so you can pick the right one.
• Search models — a search box is now available in the Settings dialog for models, VAE, text encoder, and vision models.
• Full-width image preview — click any generated image in the chat to view it larger.

Improved

• The model browser now scans model repositories more precisely, so only the files relevant to a model show up instead of unrelated files from other folders.
• Model selection is more reliable when pausing, cancelling, or resuming a download.
• The app renders more smoothly while downloading. Progress updates are now throttled so the interface stays responsive during large downloads instead of stuttering or lagging.
• After checking for new models, the app automatically refreshes the available list.
• Model and settings options are organized more cleanly.
• Session sampling overrides are now resolved before width/height, so per-session sampling changes take precedence over the model's defaults.
• The negative-prompt dialog now prefills with the session's existing prompt so editing is smoother.
• Downloaded models are excluded from the model selector when they are VAE or LoRA variants, so only full diffusion models show up for selection.

v0.1.1

Stable Release: Currently supports initial models:

• Qwen Image 2.1
• Z-Image Turbo

Change Log

Core UI & User Experience

• Implemented a complete Vue 3 chat-style interface.
• Added drag-and-drop support for reference images in the chat composer.
• Added message copy button and "save-as-image" action for generated results.
• Improved visual feedback for download progress and prompt sending.
• Updated branding, including logos, About section, and footer attributions.

Model Management

• Added comprehensive management for Diffusion, VAE, Text Encoder, and Vision (mmproj) models via settings.
• Implemented background auto-downloading for VAE and Text Encoders.
• Added support for provider-grouped directories and repo-relative paths in hub.json.
• Added "force re-download" capability for installed models.
• Filtered catalog to show only relevant diffusion entries.

Generation & Performance

• Introduced global performance/inference settings and performance profiles.
• Added support for Flash-Attention and random seed control.
• Expanded resolution options for qwen_image and z_image models.
• Hardened generation logic to prevent double-submission and sync UI states.

Bug Fixes & Stability

• Fixed startup issue where the "Starting" overlay would hang.
• Resolved model re-download loops and missing diffusion model downloads.
• Fixed pathing errors (ENOENT) occurring in packaged Windows applications.
• Fixed UI clipping during window resizing and unresponsive components after session deletion.
• Corrected various CLI argument typos and builder issues.

Build & Documentation

• Added electron-builder configuration for Windows executable packaging.
• Implemented automated CUDA runtime library downloads.
• Updated README with screenshots, a roadmap, and licensing information.