# Open-source alternatives to Topaz Gigapixel

> Open-source alternatives to Topaz Gigapixel that run locally, allow commercial use and need no paid API include waifu2x-ncnn-vulkan (photo upscaling up to 16x, recover and redefine low-quality or compressed images, upscaling art and batch upscaling of many images; MIT; runs on CPU), Real-ESRGAN (photo upscaling up to 16x, face recovery on blurry or low-res faces, upscaling art and batch upscaling of many images; BSD-3-Clause) and SeedVR2 Video Upscaler (ComfyUI) (photo upscaling up to 16x, recover and redefine low-quality or compressed images and batch upscaling of many images; Apache-2.0; GPU, ~8GB+). None of the 9 projects we track covers all 6 Topaz Gigapixel jobs, so most people combine two or more.

Topaz Gigapixel: Topaz Labs' desktop image upscaler for Windows and macOS. It enlarges photos up to 16x with a set of AI models (Standard, High Fidelity, Low Resolution, Text and Shapes, Art and CG, generative Wonder and Redefine), recovers faces, renders locally or in Topaz's cloud, batch-processes folders and plugs into Photoshop and Lightroom Classic. It does not upscale video; that is a separate Topaz product. Pricing: See pricing page (checked 2026-10-16). Facts checked 2026-10-16.

## Projects

| Project | Jobs | Runs locally | Commercial use | Paid services | Hardware |
|---|---|---|---|---|---|
| [Upscayl](https://repoquarry.dev/projects/upscayl) | photo-upscaling, batch-upscaling, art-and-text-upscaling | Yes, desktop app runs on your GPU | AGPL app; avoid the non-commercial UltraSharp and Remacri models for commercial work | None required; a separate hosted Upscayl Cloud exists | Vulkan-compatible GPU required; most integrated GPUs and CPUs do not work |
| [chaiNNer](https://repoquarry.dev/projects/chainner) | photo-upscaling, batch-upscaling, face-restoration, art-and-text-upscaling | Yes, desktop app with its own integrated Python | GPL-3.0 code; commercial use depends on each model's licence | None required | Nvidia (CUDA/TensorRT), AMD (NCNN, ROCm on Linux), Intel (NCNN), Apple Silicon (MPS); CPU fallback |
| [Final2x](https://repoquarry.dev/projects/final2x) | photo-upscaling | Yes, desktop app with a local Python backend | BSD-3-Clause covers the code; permission for commercial use of every selectable built-in or custom weight remains unverified. Check the chosen model's upstream licence | None required | Not documented; backend needs PyTorch 2.0+ |
| [QualityScaler](https://repoquarry.dev/projects/qualityscaler) | photo-upscaling, low-quality-image-recovery, batch-upscaling | Yes, fully offline | Application code is MIT; commercial-use terms of the separately downloaded model weights remain unverified in the reviewed README and licence notices | Source is free; the packaged app is sold on itch.io and Steam | 8GB RAM; any DirectX 12 GPU with 4GB VRAM or more |
| [waifu2x-ncnn-vulkan](https://repoquarry.dev/projects/waifu2x-ncnn-vulkan) | art-and-text-upscaling, photo-upscaling, low-quality-image-recovery, batch-upscaling | Yes, portable binary | Yes, MIT code and models | None required | Vulkan GPU (Intel, AMD, Nvidia, Apple Silicon); CPU mode with -g -1; smaller tiles cut GPU memory use |
| [CodeFormer](https://repoquarry.dev/projects/codeformer) | face-restoration, photo-upscaling | Yes, Python scripts on your machine | No, non-commercial licence for code and weights | None required | PyTorch 1.7.1+ with CUDA 10.1+; VRAM not stated |
| [Real-ESRGAN](https://repoquarry.dev/projects/real-esrgan) | photo-upscaling, face-restoration, art-and-text-upscaling, batch-upscaling | Yes, portable executable or Python script | Yes, BSD-3-Clause; the optional --face\_enhance step uses GFPGAN, which includes non-commercial parts | None required | Portable build runs on Intel, AMD or Nvidia GPUs via Vulkan; tile option reduces GPU memory |
| [GFPGAN](https://repoquarry.dev/projects/gfpgan) | face-restoration, photo-upscaling | Yes, Python script on your machine | Unclear; bundled third-party code includes NVIDIA and DFDNet non-commercial terms | None required | PyTorch; NVIDIA GPU with CUDA optional; the clean version needs no custom CUDA extensions |
| [SeedVR2 Video Upscaler (ComfyUI)](https://repoquarry.dev/projects/seedvr2-comfyui) | photo-upscaling, low-quality-image-recovery, batch-upscaling | Yes, inside ComfyUI or as a standalone CLI | Yes, Apache-2.0 code and official weights; third-party GGUF conversions may differ | None required | 8GB VRAM or less with GGUF Q4 and BlockSwap; 12-16GB with FP8; 24GB+ for FP16 |

## Common questions

### Is there a free, open-source alternative to Topaz Gigapixel?

Yes. chaiNNer, Upscayl, waifu2x-ncnn-vulkan, Real-ESRGAN and QualityScaler are free to download and run on your own computer without paid API keys. Each covers only part of what Topaz Gigapixel does, so check which job you need.

### Can I do photo upscaling up to 16x locally without a Topaz Gigapixel subscription?

Yes, with Upscayl (AGPL-3.0), chaiNNer (GPL-3.0) and waifu2x-ncnn-vulkan (MIT). Check the hardware notes and licence conditions on each page before you start.

### Can I do face recovery on blurry or low-res faces locally without a Topaz Gigapixel subscription?

Yes, with chaiNNer (GPL-3.0), Real-ESRGAN (BSD-3-Clause) and GFPGAN (Apache-2.0). Check the hardware notes and licence conditions on each page before you start.

### Which Topaz Gigapixel alternatives can I use commercially?

Confirmed: Real-ESRGAN, SeedVR2 Video Upscaler (ComfyUI) and waifu2x-ncnn-vulkan. With conditions: chaiNNer, GFPGAN and Upscayl. Not for commercial use: CodeFormer, usually because the model weights are non-commercial. This is not legal advice; read each licence.

### Does any open-source project replace all of Topaz Gigapixel?

Not as of 2026-10-16. Photo upscaling up to 16x, face recovery on blurry or low-res faces, recover and redefine low-quality or compressed images, upscaling art, batch upscaling of many images and Photoshop and Lightroom plugin are covered by different projects.

---
Source: RepoQuarry (https://repoquarry.dev/alternatives/topaz-gigapixel). Data licence: CC BY 4.0. Quotes are from project documentation and belong to their authors.