How is this different from a standard sharpen filter?

Instead of boosting existing edges, LTX 2.3 with the Refocus IC LoRA uses a learned prior to reconstruct plausible detail. This can recover mid‑frequency structure that simple unsharp masks miss, but overly strong settings can introduce artifacts—keep the LoRA weight moderate.

Does the workflow handle audio?

No. VHS_VideoCombine outputs a video from processed frames only. Export your enhanced clip and reattach the original audio in your NLE or a separate muxing step.

What footage benefits most?

Clips that are slightly soft or marginally out of focus. Extreme defocus or heavy motion blur won’t fully recover—try small increases in LoRA strength and evaluate on a short segment before processing the full video.

What performance should I expect?

Speed depends on resolution, clip length, and GPU VRAM. Processing per-frame with a generative model is compute-heavy; test on a short frame range first and consider batching or downscaling for faster previews.

Refocus LTX 2.3 IC LoRA - ComfyUI Workflow

This workflow restores soft or slightly out‑of‑focus footage using the Refocus IC LoRA on top of the LTX 2.3 model. It ingests your clip with VHS_LoadVideo, processes each frame through a custom LTX 2.3 + LoRA processor (node UUID: 2e37ecc9-770a-4ac2-972c-e953901f49d3), and then reassembles the enhanced frames back into a video with VHS_VideoCombine. The Refocus IC LoRA biases the LTX 2.3 backbone toward reconstructing fine edges and mid‑frequency detail that typical sharpening filters can’t recover, helping pull subjects back into clarity while minimizing halos and ringing.

Technically, the pipeline works frame‑wise: VHS_LoadVideo decodes your source and outputs frame tensors; the custom node loads the LTX 2.3 base and applies the Refocus IC LoRA weights to guide inference toward crisp, natural detail; and VHS_VideoCombine stitches the processed frames into a finished clip at your chosen frame rate and quality. Because this is model‑driven enhancement rather than a simple unsharp mask, it can plausibly restore subtle textures and edge definition in footage that would otherwise be unusable. Keep LoRA strength conservative for realism, and iterate to balance sharpness against potential artifacts.

API

Use this workflow from code

Every Comfy workflow is a JSON graph. The payload below is this workflow, exactly as ComfyUI runs it — fetch it from the URL, keep it in version control, or load it in ComfyUI and run it node by node.

4ca799182440.json

Fetching workflow JSON…

Run it from TypeScript or Python with the Comfy SDK. The same code targets Comfy Cloud or a ComfyUI you host yourself — only the base URL changes.

// Install (beta)
npm i @comfyorg/sdk

// Run this workflow (TypeScript)
import { Comfy } from "@comfyorg/sdk";

const client = new Comfy({ apiKey: "comfyui-..." });
const wf = await client.workflows.fromFile("workflow_api.json");
const job = await client.run(wf);
await job.getOutputs("<output-node-id>")[0].toFile("output.png");

The SDK takes a workflow in API format: open this workflow in ComfyUI and use File → Export Workflow (API).

Comfy Cloud API access requires a plan with an API key.

SDK docs Get an API key

FAQ

Frequently Asked Questions