SeedVR2 · Image Upscaler

ABOUT THE WORKFLOW

Upscale an image
Upload an image and get a sharper, higher-resolution version back. The model rebuilds detail, reduces noise, and raises the resolution in a single step. No prompt needed, since it works from the image alone.

Partner node. This workflow calls an external API, so each run uses credits from your API wallet. No API key needed. Floyo handles the connection.

Model

• SeedVR2 by ByteDance Seed Team. A one-step diffusion restoration model, accepted at ICLR 2026, that upscales and cleans up an image in a single pass, rebuilding natural detail rather than only sharpening. Open source under Apache 2.0, and this workflow runs a hosted version.

HOW IT WORKS

Step 1. Upload your image
The image you want to upscale. Lower-resolution inputs gain the most.
Works great with: photos · old or compressed images · AI-generated images · screenshots

Step 2. Set how much to upscale (optional)
Keep the 2x factor, or switch to a target resolution like 1080p for a fixed output size.

Step 3. Hit run and download
The model rebuilds detail and raises the resolution in one step, then saves the result.
Ready for: print · web · social · client delivery

First time? Leave every setting as-is. The defaults (2x · factor mode · low noise) are the right starting point for almost everyone.

RECOMMENDED SETTINGS

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

• Standard use (most people) — 2x factor, low noise. The right starting point for almost everyone.

• Want a fixed output size — switch the mode to target resolution and pick a size like 1080p, instead of multiplying by a factor.

• Faithful upscaling — keep the noise scale low, around the default. The model rebuilds detail without inventing much.

• Want more invented detail — raise the noise scale in small steps. It adds detail but can start to hallucinate.

• Cleaning up a rough image — a bigger upscale factor gives more room to rebuild, though heavily degraded inputs may not fully clean up.

• Want the exact same result again — set a fixed seed in place of random. Random gives you a fresh take on each run.

Input image: This workflow needs no prompt, so the input image is what matters. A lower-resolution or slightly soft image gains the most, while an already-sharp image has little to gain. Heavily compressed or damaged images can only be rebuilt so far, so start from the cleanest source you have.

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 Upscaling
Turn a small or soft photo into a print-ready, higher-resolution version.

🎨 AI Image Cleanup
Sharpen and add detail to AI-generated images before delivery.

🗂️ Archive Restoration
Bring old, low-resolution, or compressed images up to a usable size.

🛍️ Product & Catalog
Raise product shots to a consistent resolution for a store or catalogue.

WHAT WORKS BEST / WHAT TO AVOID

✅ Works great

• Low-to-mid resolution photos

• Slightly soft or compressed images

• AI-generated images that need detail

• Clear subjects with natural texture

⚠️ May produce softer results

• Heavily degraded or noisy inputs

• Images that are already sharp

• Extreme target output sizes

• Small text and faces in the frame

FAQ

What is SeedVR2?
SeedVR2 is a one-step diffusion restoration model from ByteDance's Seed Team, accepted at ICLR 2026. It upscales and restores an image in a single pass, rebuilding natural detail while reducing noise, and it works on video too. The model is open source under Apache 2.0, and this workflow runs a hosted version of it.

How is SeedVR2 different from other upscalers?
SeedVR2 restores an image in one step, where many diffusion upscalers need dozens, so it is fast for the quality. It pairs a diffusion transformer with adversarial training to rebuild believable detail rather than only sharpening edges, which puts it on par with commercial upscalers. The trade-off is that on lightly degraded inputs it can over-sharpen or invent detail.

Do I need a prompt to use SeedVR2?
No. SeedVR2 works from the image alone, with no text prompt. You upload a picture and the model rebuilds detail and raises the resolution, so the only real input is the quality of the image you start from.

What does the noise scale setting do?
Noise scale controls how much new detail the model invents. A low value, around the default, keeps the result faithful to your original, while a higher value adds more detail and can start to hallucinate textures. Keep it low for restoration and raise it only when you want the model to fill in more.

Is SeedVR2 free and open source?
The SeedVR2 model is open source under the Apache 2.0 license, so it is free to run yourself if you have the hardware. This workflow uses a hosted API version instead, which saves you the setup and the GPU, so each run draws credits from your API wallet.

Can I use the upscaled images commercially?
Yes. SeedVR2 is Apache 2.0, which permits commercial use, and an upscaled image belongs to you as long as you hold the rights to the original you uploaded. The hosted version in this workflow is a paid service, but it does not add any extra license restriction on the result.

How to run SeedVR2 online?
You can run SeedVR2 online through Floyo. No installation, no setup, no API key to wire up. Open the workflow in your browser, upload an image, 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 upscales a shot 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 an image and run it. No prompt needed.

→ Launch Workflow, Free

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