A workflow combining two-stage F3K restoration, the Quality LoRA, and optional upscaling to rebuild image detail and produce high-resolution results. It needs to work together with the F3K LoRA.

like this

✨HighQuality for F3K✨private version

✨MoreReal✨for MOSS-image

First, it looks like this

How It Works

• Stage 1 — Restore: Rebuild the image at a lower resolution to improve clarity and establish the overall structure.

• Stage 2 — Refine: Upscale and re-encode the first-stage result, then refine it using the later sampling steps.

• Optional enhancement: Enable DLSS5 and VOSR2 as needed. Use the built-in A/B comparison to inspect the results.

The example uses 30 total steps, a refinement start step of 18, and a LoRA strength of 0.9.

With the same total step count, starting refinement earlier produces stronger changes; starting later preserves more of the first-stage result. Use these settings as a starting point and adjust for your image.

Setup

Model download pages and custom-node GitHub links are included on the right side of the workflow.

Final dimensions depend on the image’s aspect ratio and enabled upscale stages. The example shown reaches 7552 × 8832. Fine details are AI-reconstructed, so check faces and textures before using the result.

OK! That's all! I hope you like it!

If you like my sharing and want to get more content and resources related to AI image generation, welcome to join my private community.👇👇👇

NO8D-AI-Workshop

Join Us

🫡May the AI-power be with you, see you soon!🫡

Disclaimer: My model or workflow is solely for learning and communication purposes and shall not be used for any commercial activities. All responsibilities for the generated content, including its legality, ethical implications, or potential misuse, are borne entirely by the user. By using this model or workflow, you agree to this principle. Please use it responsibly within the scope of local laws.