FLUX.1 Dev · Image Upscaler

ABOUT THE WORKFLOW

Double the Resolution of Any Image Upload a picture and hit run. The 4x-UltraSharp model scales it up, then FLUX.1 Dev refines each tile at low denoise to sharpen detail without changing the content. The tiles are stitched back together and saved at twice the original size.

Model

• FLUX.1 Dev by Black Forest Labs. Released August 2024. A 12 billion parameter rectified flow transformer. Open weights under the FLUX.1 Dev Non-Commercial License. Used here as the tile refiner, not as a generator.

• 4x-UltraSharp upscale model. A pretrained super-resolution network that does the initial 4x resize before FLUX.1 Dev refines each tile back to the target 2x factor.

HOW IT WORKS

Step 1. Upload your image Any size. The output will be twice as large on each side. Works great with: photos · renders · AI-generated images · scans

Step 2. Add a prompt (optional) Leave it empty for a clean upscale. Add a short description of the image to guide the refinement if you want the model to lean into specific detail.

Step 3. Hit run and download The image is scaled up, split into 1920 x 1080 tiles, each tile is refined in 20 passes at low denoise, and the tiles are stitched together. The result is saved under Image_Upscaler/Image. Ready for: Photoshop · Lightroom · print · any editor

First time? Leave every setting as-is. The defaults (2x upscale · 0.15 denoise · 20 steps · guidance 3.5 · random seed) are the right starting point for almost everyone.

RECOMMENDED SETTINGS

Quick-start guide. Find the goal that matches yours and copy the settings.

• Standard upscale (most people) — 2x · 0.15 denoise · 20 steps · guidance 3.5 · random seed. The right starting point for almost everyone.

• The upscale looks too soft — Raise denoise from 0.15 toward 0.25. The model redraws more of each tile, adding sharper detail. Go too high and the content starts changing.

• The upscale changed the content — Lower denoise below 0.15. At values near 0.05 the model barely touches the image, producing a cleaner but less detailed scale.

• Tile seams are visible — Keep denoise low. High denoise makes each tile diverge from its neighbours, and the stitch line shows. Below 0.2 seams are rare.

• The run is taking a long time — Large images at 2x create many tiles, and each tile runs 20 passes through a 12 billion parameter model. Lower steps to 10 for a faster result at some cost to refinement.

• Want to guide the refinement — Add a short prompt describing the image. "Interior of a baroque palace, gold leaf, marble columns" steers the model toward the right kind of detail in each tile.

• Want a larger upscale — Raise the upscale factor above 2. More tiles are created, the run takes longer, and seam risk increases with higher denoise.

• Reproduce a result you liked — Set a fixed seed. The same image, prompt, and seed return the same upscale.

Input image: The cleaner and sharper the input, the better the upscale. Heavily compressed JPEGs, screenshots with interface chrome, and images with text overlays produce weaker results because the model amplifies whatever it finds.

LEARN

📹 Videos

• Intro to Floyo

• ComfyUI 101 Free Course ft. Sebastian Kamph

• Floyo 101 for Team Collaboration

✨ Quick links

• Help docs

• LoRA Training guide

• Discover more workflows

USE CASES

📸 Photo Enlargement Double the resolution of a photo for print, large-format display, or high-DPI screens.

🎨 AI Art Upscaling Sharpen and enlarge AI-generated images that were produced at lower resolution.

🖼️ Scan and Archive Enhancement Scale up scanned artwork, documents, or old photographs with added sharpness.

🛍️ Product Image Prep Enlarge product shots to meet marketplace resolution requirements without reshooting.

WHAT WORKS BEST / WHAT TO AVOID

✅ Works great

• Clean, well-exposed source images

• Photos and renders at any starting resolution

• Denoise between 0.1 and 0.2 for faithful upscales

• Adding a prompt to guide detail recovery

⚠️ May produce softer results

• Heavily compressed JPEGs with visible artifacts

• Denoise above 0.3, where tiles start to diverge

• Images with readable text (text may soften or shift)

• Screenshots with interface elements or overlays

FAQ

What is FLUX.1 Dev? FLUX.1 Dev is a 12 billion parameter rectified flow transformer from Black Forest Labs, released August 2024. In this workflow it is used as a tile refiner, not as a generator. Each tile of the upscaled image is passed through FLUX.1 Dev at low denoise to sharpen detail without changing the content.

What is 4x-UltraSharp? It is a pretrained super-resolution network that handles the initial resize. It scales the image up by 4x in one pass, and then UltimateSDUpscale tiles it back to the target factor (2x by default) while FLUX.1 Dev refines each tile. UltraSharp handles geometry and structure; FLUX.1 Dev handles texture and fine detail.

What does denoise control? How much the model redraws each tile. At 0.15 (the default) it sharpens and adds fine detail while keeping the original content intact. Raising it toward 0.3 adds more detail but risks changing what is in the tile. Lowering it toward 0.05 produces a cleaner scale with less added detail.

Why does the workflow split the image into tiles? FLUX.1 Dev cannot process a high-resolution image in one pass. Tiling splits the upscaled image into manageable pieces (1920 x 1080 by default), refines each one independently, and stitches them back together. The tile overlap and linear blending mode reduce visible seams.

Is FLUX.1 Dev free for commercial use? FLUX.1 Dev is released under the FLUX.1 Dev Non-Commercial License, which does not permit commercial use. A separate FLUX.1 Pro licence from Black Forest Labs covers commercial applications. The 4x-UltraSharp model is a community release with its own terms. Check both before using output in paid work.

How long does the upscale take? It depends on the input size. A small image produces fewer tiles and finishes in seconds. A large image at 2x creates many tiles, and each tile runs 20 passes through a 12 billion parameter model. Expect longer runs on big inputs.

How to run FLUX.1 Dev upscaler online? You can run FLUX.1 Dev upscaler online through Floyo. No installation, no setup, no model downloads. Open the workflow in your browser, upload a picture, and hit run. Free to try.

WHY FLOYO?

Floyo is the only platform with team collaboration for ComfyUI in the browser. You run workflows with no install. You share run history, assets, and models across your team. You pay only when you generate. Floyo supports open-source and closed-source models.

A designer runs an edit and likes the result. A teammate opens that exact run from shared history and keeps going. No file handoffs. No version confusion.

For studios and enterprise teams, Floyo adds private workspaces, pooled resources, and a team usage dashboard. Other ComfyUI cloud tools run for one person at a time. Floyo runs for the whole team, with transparent per-generation costs.

Ready to try it? Upload any image and run it. The result comes back at double the resolution.

→ Launch Workflow, Free

Questions? Watch the free course or check the FAQ above.