Upload your image. Click Run. Get a 2x result.

Easy2x is a free, simple ComfyUI workflow by ClearFlowAI. Double the width and height of one image using a lightweight SPAN upscaler: 512 × 512 → 1024 × 1024.

4 built-in processing nodes · 1 small model · No prompt · No API key · No custom-node installation

First run — 6 steps

1. Download 2xNomosUni_span_multijpg_ldl.pth (about 8.95 MB).

2. Put it in ComfyUI/models/upscale_models/. Refresh or restart ComfyUI.

3. Drag ClearFlowAI_Easy2x_v1.0.json onto the canvas. The optional .api.json is for API users.

4. Upload/select your local image in node 1. The sample filename in the workflow is a placeholder until imported.

5. Select the exact 2x model in node 2.

6. Click Run / Queue Prompt. Node 4 previews the PNG. Find saved files in your configured output folder under ClearFlowAI/Easy2x_*.png.

The START_HERE ZIP includes the canvas workflow, optional API graph, guide, sample input, actual sample output and test report. It contains no model weights. Instructions are also inside the canvas.

Actual test

Original synthetic still-life input: 1254 × 1254 → 2508 × 2508. The uploaded result was processed through this workflow. Tested backend: ComfyUI 0.31.1, PyTorch 2.12.1, Apple MPS, 16 GB unified memory. Execution, polling and saving took 6.22 seconds on this machine; your timing and memory needs may differ.

Help and limits

Model missing? Check the filename/folder, refresh and re-select. Image missing? Upload/select a file in node 1. Out of memory? Try a smaller input.

Desktop compatibility: frontend 1.48.7 with backend 0.31.1 showed a media-picker warning despite successful backend and UI-queued execution. Use matching current frontend/backend versions and re-upload/re-select the image. The included API graph runs on the tested backend.

This estimates texture; it cannot recover true lost detail or reliably restore unreadable text. It is not diffusion redraw or dedicated face restoration. RGB output; transparency is not preserved. Other upscaler models may change the scale factor.

Model credit

NomosUni SPAN multijpg LDL by Helaman, based on the NomosUni family by Philip Hofmann (Phips). Model source and CC BY 4.0 license. ClearFlowAI assembled the workflow and did not train the model. Model licensing applies separately.

Need help? Comment with your ComfyUI version, OS, input dimensions and exact error. Follow ClearFlowAI for workflows and LoRAs with a clear purpose and clear instructions.