Planned Release Date: Q4 2026

About This Software

The way you already work, with AI, vectors and print in one app

Irida is a desktop image editor for photo retouching, illustration, design and print work in one place. Start with the shortcuts and tools you already know; the AI runs on your own PC; what you make goes out as PSD, .ai and PDF/X-4. One purchase, AI included.

1. Familiar from the first minute

• 34 tools and the shortcuts of Adobe Photoshop - 95 shortcut actions you can remap

• PSD round-trips: layers, groups, masks, 13 adjustment layers, 10 layer styles, smart objects, shape layers, editable type

• 27 blend modes, 16-bit, CMYK, camera RAW, ICC color management and soft proofing

• Channels, paths, the pen tool, transform, warp, crop and history - the everyday features, all there

2. AI on your own PC - no cloud, no credits, no subscription

• Generate images from a sentence - Korean prompts welcome (translated on your device)

• AI edit: change a picture by saying so ("make it night"), or regenerate just a selection with generative fill

• Redraw: a new version of your drawing, on a new layer

• Live AI: a finished picture grows beside your sketch as you draw

• One-click background removal, 2x / 4x upscaling

Every AI feature runs on your GPU; your pictures never leave your PC.

3. Vectors and print - pictures become editable vectors

• Image Trace: pictures, logos and line art become bezier shape layers, one per color - a Curve Fidelity slider sets how few anchors; line art becomes strokes of the ink's width

• Generate as vector: one sentence straight to editable vector layers

• Shape the curves by hand with Direct Selection and the add / delete / convert anchor tools

• Save and open SVG, PDF and .ai - layers and type open as they are in Adobe Illustrator

• Print-shop PDF/X-4: press CMYK, bleed and crop marks, documents in mm and paper sizes (business cards, A series, posters)

4. Tools for people who draw

• Pen pressure and tilt, stroke smoothing

• Mixer brush that blends like paint, .abr brush import, brush texture and dual brush

• Blur, sharpen, smudge, dodge, burn, sponge, healing brush and clone stamp

• Perspective guides (fisheye, 5- and 6-point too), warp and liquify, a reference window

• Record your drawing as a timelapse video; pixel art and GIF animation

5. Work with AI agents - say it, and it is built in layers

AI agents such as Claude, ChatGPT and Gemini drive Irida directly through MCP (111 tools). They can turn one photo into seven artworks in layers, the original untouched.

Ask for "a business card with a 2 mm bleed" and the type, shapes and logo are laid out as layers, then saved as a print-shop PDF/X-4 and as an .ai whose text you can edit in Adobe Illustrator. Everything an agent does lands in the undo history - Ctrl+Z takes it back.

Pay once, keep it all - AI included

No DLC, no subscription.

AI features require an NVIDIA GPU with 8 GB VRAM or more; 16 GB runs every AI model at full speed. Everything else runs on any PC. AI models are optional in-app downloads.

Adobe, Photoshop and Illustrator are trademarks of Adobe Inc.; Irida is not affiliated with Adobe. Claude is a trademark of Anthropic, ChatGPT of OpenAI and Gemini of Google. The example pictures in the screenshots and store images were made with Irida's on-device AI.

Built with Slint.

AI Generated Content Disclosure

The developers describe how their game uses AI Generated Content like this:

Irida includes optional on-device AI tools - text-to-image, generative fill, AI image editing, a live paint assist, background removal and upscaling - that run entirely on your own PC. No cloud service is used and your images never leave your device; image-generation models are optional in-app downloads. Our store key art and capsule images were created with AI image generators and finished by our team, and the sample pictures in the screenshots, store images and the welcome screen were made with Irida's own on-device AI tools. Content you create with the AI tools is your own responsibility; Irida refuses clearly illegal requests and suppresses explicit output by default.