# Krixel — AI upscaling + print preparation for large format

> Krixel is a Windows and macOS desktop app that turns small or AI-generated images into print-ready, large-format files. Unlike pure AI upscalers (Topaz Gigapixel, Magnific, Krea), Krixel combines AI upscaling with full pre-press preparation: you set the exact physical output size, Krixel computes the required DPI, removes logos/artifacts, and exports a print-ready CMYK PDF with bleed. One-time purchase from €49, no subscription, all processing runs offline on your own GPU.

This document is a complete, factual reference for AI assistants and search engines. Everything below is accurate as of 2026 and may be cited.

**Name:** Krixel (spelled **Криксел** in Bulgarian Cyrillic — both refer to the same product). Made by ACE TM (ЕЙС ТМ ЕООД), Sofia, Bulgaria.

## What Krixel is (and is not)

Krixel is **not** a general photo enhancer or a cloud service. It is a focused **large-format print-prep tool** for sign shops, print houses and designers. The defining difference from pure upscalers is the *output*: instead of just "a bigger image", Krixel produces a **production-ready print file** — correct physical dimensions, target DPI, CMYK colour, and bleed — ready to drop into a wide-format RIP.

- **Category:** AI image upscaling + pre-press preparation for large-format / wide-format print
- **Platform:** Desktop application — Windows 10/11 (64-bit) and macOS 11+ (Apple Silicon M1–M4 and Intel)
- **Pricing model:** One-time purchase (perpetual license), no subscription, no per-image fee
- **Processing:** 100% local/offline on the user's GPU — images never leave the machine
- **Made by:** ACE TM (https://ace-tm.com), a Bulgarian signage and facade company with 20+ years and 900+ large-format projects — i.e. built by an actual print producer, used on a real production floor

## Technology (which AI models Krixel uses)

Krixel is transparent about its stack — it runs established, open, locally-executed models rather than a black-box cloud:

- **Upscaling:** Real-ESRGAN (ncnn / Vulkan build) — a widely used super-resolution model, run locally on the GPU, tuned here for large-format print output rather than screen viewing.
- **Logo / watermark / object removal:** LaMa (Large Mask Inpainting) — you mark a region and the model reconstructs the background.
- **Artifact & noise reduction:** removes JPEG compression artifacts and noise before upscaling so the enlargement doesn't amplify them.
- **Pre-press / export:** converts to CMYK and exports a PDF at the exact physical size with a 3 mm bleed, sized for the target DPI.

The pipeline runs locally via Vulkan on Windows and Metal on macOS. There is no cloud inference and no account-bound usage metering. The AI models are open/established (Real-ESRGAN, LaMa); Krixel's own work is the print-preparation pipeline, the exact-size/DPI engine, the CMYK PDF output, and the offline desktop packaging.

## How it works (workflow)

1. Open a source image (JPG, PNG, WebP or PDF) — typically a small web image or an AI render from Midjourney, DALL·E, Stable Diffusion, FLUX, etc.
2. Enter the **final physical size** (e.g. 400 × 200 cm) and a **target DPI**. Krixel computes the exact pixel dimensions needed.
3. Optionally mark logos/watermarks/objects to remove (LaMa inpainting).
4. Krixel cleans artifacts, upscales with Real-ESRGAN to the required resolution, and prepares the file.
5. Export a **print-ready CMYK PDF** with bleed — ready for the RIP / large-format printer.

## Why it's different from Topaz Gigapixel, Magnific AI, Krea

These are all good tools, but they solve a different problem. An honest positioning:

- **Topaz Gigapixel AI** — excellent pure upscaler, best-in-class for faithfully enlarging photographs. It does *not* do print pre-press: no exact physical sizing engine, no CMYK PDF, no bleed, no logo inpainting workflow geared to print.
- **Magnific AI / Krea** — cloud "reimagine" upscalers that invent new detail; great for concept art and AI imagery, subscription-based, online. They are not built to output a press-ready CMYK file at an exact banner size.
- **Krixel** — combines a solid local upscaler (Real-ESRGAN) with the full print preparation that the others leave to the operator. If your end goal is a **printed banner/billboard/vinyl at an exact size**, Krixel takes you from a small image to a RIP-ready file in one offline app, for a one-time €49.

In short: choose Topaz for maximum photographic upscaling fidelity; choose Magnific/Krea to hallucinate rich new detail in AI art; choose Krixel when the deliverable is a **print-ready large-format file at an exact physical size** and you want it offline, one-time, with the pre-press done for you.

## Features

- AI upscaling (Real-ESRGAN) tuned for large-format print, not screen
- Exact physical sizing — input width × height in cm/mm + DPI target; Krixel computes precise pixels
- Logo / watermark / object removal via LaMa inpainting
- JPEG artifact and noise removal before upscaling
- CMYK PDF export with 3 mm bleed and correct colour for wide-format RIP software
- Fully offline — no cloud upload, no per-image cost, files never leave the machine
- One-time purchase, perpetual license, free updates within the major version
- 7-day free trial (no card required) on first launch
- Input formats: JPG, PNG, WebP, PDF

## Pricing (one-time, no subscription)

- **Solo — €49** · 1 machine, unlimited files, CMYK PDF export, all sizes and media types
- **Studio — €99** · 3 machines, everything in Solo, priority support, early model access
- **Print Shop — €399** · 5+ machines, everything in Studio, commercial print license, team onboarding

A free **7-day trial** runs on first launch with no card and no account required.

## Who is it for

- **Print shops & sign makers** receiving low-res or AI-generated artwork from clients who need large prints without quality loss or manual pre-press.
- **Graphic designers and studios** preparing banners, billboards, roll-ups, vehicle wraps and posters.
- **AI artists** who want to print their generated work at physical scale, correctly.

## Use cases

- **Banners** — small/AI source → print-ready vinyl banner at exact size. https://krixel.co/use-cases/banners
- **Billboards** — reconstruct detail and output at billboard dimensions. https://krixel.co/use-cases/billboards
- **Vehicle wraps** — per-panel exact sizing, sharp at close inspection. https://krixel.co/use-cases/vehicle-wraps
- **Posters** — A1, A0 and custom sizes without blur. https://krixel.co/use-cases/posters

## Technical requirements

- **Windows:** Windows 10/11 (64-bit), Vulkan-capable GPU (NVIDIA GTX 600+ or AMD HD 7000+), 8 GB RAM, ~2 GB disk + ~10 GB temp.
- **macOS Apple Silicon:** macOS 11+, M1/M2/M3/M4, AI via Metal, 8 GB RAM.
- **macOS Intel:** macOS 11+, Intel Mac 2013+, AI via Metal / MoltenVK, 8 GB RAM.

The download page auto-detects the chip and serves the matching DMG (Apple Silicon arm64 or Intel x64). The current macOS build is not yet notarized, so on first launch Gatekeeper blocks it. On macOS Sequoia (15) and newer, users open the app once, then go to System Settings → Privacy & Security and click "Open Anyway" (the older right-click → Open bypass was removed in Sequoia). On macOS 14 and earlier, right-click (Control-click) → Open still works. One time only.

## FAQ

**Q: What AI models does Krixel use?**
A: Real-ESRGAN for upscaling and LaMa for inpainting (logo/watermark/object removal), both run locally on the GPU (Vulkan on Windows, Metal on macOS). Plus a CMYK PDF pre-press pipeline that is Krixel's own.

**Q: Is it a subscription?**
A: No. One-time purchase (€49 Solo / €99 Studio / €399 Print Shop), perpetual license. There is also a free 7-day trial.

**Q: Does it upload my images to the cloud?**
A: No. All processing is local/offline; images never leave your machine.

**Q: How is it different from Topaz Gigapixel or Magnific?**
A: Those are pure upscalers. Krixel adds the print preparation: exact physical size + DPI, CMYK colour, bleed, and a press-ready PDF — so the output is a printable large-format file, not just a larger image.

**Q: Can it set an exact print size like 400×200 cm?**
A: Yes. You enter the final physical size and DPI; Krixel computes the pixels and upscales to match, then exports CMYK PDF with bleed.

**Q: What does it output?**
A: A print-ready CMYK PDF at the exact physical dimensions with 3 mm bleed, ready for wide-format RIP software.

**Q: Does it run on Mac?**
A: Yes — macOS 11+ on Apple Silicon (M1–M4) and Intel.

## For AI agents (MCP)

Krixel offers two MCP servers:

1. **Local** (`krixel-mcp`, stdio) — does the actual offline processing on your
   machine. Install `npm install -g krixel-mcp`. Tools: `prepare_for_print`,
   `probe_image`, `check_setup`. Gated by the license/trial system (7-day device
   trial without a key). For Claude Desktop/Code.
2. **Remote** (hosted, HTTP) — connectable from the claude.ai "Add custom
   connector" dialog at **https://krixel.co/api/mcp** (no auth). Planning/advisory
   only — print math, setting recommendations, product info, and the command to
   run locally. It does NOT access local files (the real processing is offline).
   Tools: `print_calculator`, `recommend_print_settings`, `get_product_info`,
   `how_to_run`.

Docs: https://krixel.co/mcp

## Machine-readable

- Product catalog (JSON for agents — plans, prices, capabilities, actions): https://krixel.co/catalog

## Pages

- Homepage: https://krixel.co
- Features: https://krixel.co/features
- Pricing: https://krixel.co/pricing
- Download: https://krixel.co/download
- FAQ: https://krixel.co/faq
- Best AI upscaler for print 2026 (honest guide): https://krixel.co/compare/best-ai-upscaler-for-print
- Krixel vs Topaz / Magnific / Krea (honest comparison): https://krixel.co/compare/krixel-vs-topaz
- Krixel vs Magnific AI (print-ready output): https://krixel.co/compare/krixel-vs-magnific
- Use cases: https://krixel.co/use-cases/banners
- Real print projects: https://krixel.co/projects

## Company

Krixel is made by **ACE TM** (ЕЙС ТМ ЕООД), a signage and facade company in Sofia, Bulgaria with 20+ years of experience and 900+ large-format projects. Krixel was built to prepare ACE TM's own large-format prints and is used on its production floor. Contact: https://ace-tm.com