FLUX + SeedVR2 — Image Upscaling Workflow
A carefully tuned ComfyUI workflow for upscaling generated images
A ready-to-run workflow, supporting models, and visual examples — all in one place.
✨ Overview
This is the best image-upscaling setup I have found for generated images — a solution I spent a long time tuning and perfecting.
The repository includes a ready-to-run ComfyUI workflow, model resources, and example outputs. You must install the custom SeedVR nodes to use the workflow.
Important: This workflow upscales an image you provide. It does not generate an image from scratch.
🖼️ Upscaling Examples
🔍 Original Images
The original images, before upscaling, are available here:
Browse the original images on Hugging Face
📦 Required Models
1. FLUX.1 diffusion model
You can use the standard FLUX.1-dev model, or the FLUX model files provided here:
FLUX NSFW Unlocked — model files
2. FLUX.1 text encoder (CLIP)
FLUX NSFW Unlocked — CLIP files
3. SeedVR2 upscaling models
SeedVR2 ComfyUI models by numz
Place the downloaded SeedVR2 model files in:
ComfyUI/models/SEEDVR2/
If you install the nodes through ComfyUI's Manager or download manager, the models may be downloaded automatically the first time you run the workflow.
🧩 Custom Nodes
Install the custom SeedVR nodes before loading the workflow. If you use a ComfyUI manager, install the relevant SeedVR node package there and follow its setup instructions. https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler
Install the StarNodes use a ComfyUI manager or https://github.com/Starnodes2024/ComfyUI_StarNodes Installation is not required; if you do not need layered Photoshop files as the output, simply remove the StarNodes node from the workflow.
🛠️ Workflow
The repository includes the workflow image:
🔞 SFW & NSFW Compatibility
The upscaler can process both SFW and NSFW images. The workflow itself does not inherently generate NSFW content: it takes an existing image and upscales it.
FLUX may be particularly useful for upscaling certain types of NSFW images, but results depend on the source image, model files, and workflow settings. Always follow applicable laws, model terms, and platform rules.
📜 Licenses & Attribution
The repository metadata lists Apache-2.0, but the workflow depends on third-party models and custom nodes that may have separate license terms.
- FLUX.1-dev: Black Forest Labs — FLUX.1-dev
- SeedVR2 (ComfyUI): numz/SeedVR2_comfyUI
- SeedVR2: Comfy-Org/SeedVR2
Please check the current license and usage terms published by each model author before downloading, redistributing, or using the models. The repository's license metadata does not override third-party model licenses.
🙏 Credits
- Black Forest Labs — FLUX.1-dev.
- Comfy-Org — SeedVR2.
- numz — SeedVR2 ComfyUI model resources.
- The authors of the SeedVR custom nodes and the ComfyUI workflows that informed this setup.
Upscale your images. Keep creating.