> ## Documentation Index
> Fetch the complete documentation index at: https://docs.comfy.org/llms.txt
> Use this file to discover all available pages before exploring further.

# VideoDeflicker - ComfyUI Built-in Node Documentation

> Removes temporal flicker from a processed video, so each frame keeps its own detail while following the motion of its neighbours.

Removes temporal flicker from a processed video, so each frame keeps its own detail while following the motion of its neighbours. It is meant for footage that pulses after being upscaled or processed frame by frame: connect the optical flow model and the flickering frames, and optionally a stable copy of the source as a guide.

## Inputs

| Parameter | Description | Data Type | Required | Range |
| - | - | - | - | - |
| `optical_flow` | Optical flow model from the Load Optical Flow Model node (torchvision RAFT-large). | OPTICAL\_FLOW | Yes | - |
| `images` | Flickering frames to stabilise, for example a video that was upscaled or processed frame by frame. | IMAGE | Yes | - |
| `guide` | Stable source frames to follow the motion of, at any resolution. Defaults to `images`, which removes less flicker. Its motion-compensated frame differences also weight the smoothing, so a guide that changes in ways the flow cannot explain smooths less even when the motion estimate is right. | IMAGE | No | - |
| `strength` | How strongly each frame follows its neighbours (default: 1.0). Higher values remove more flicker and soften more fine detail; 0 returns the frames unchanged. | FLOAT | Yes | 0.0 to 100.0 (step 0.1) |
| `direction` | Direction of the pass. `bidirectional` averages a forward and a backward pass, which avoids trailing behind the motion; `forward` runs a single pass. (default: `bidirectional`) | COMBO | Yes | `"bidirectional"` `"forward"` |
| `threshold` | Changes larger than this between a frame and its motion-compensated neighbour are treated as real and smoothed less (default: 0.1). 0 smooths every change alike. | FLOAT | Yes | 0.0 to 1.0 (step 0.01) |

**Notes:**

* The optical flow runs on a downscaled copy of the guide, so it does not need to match the size of `images`; the guide's pixels are not blended into the output.
* Frames past the shorter of `images` and `guide` are not deflickered. With a shorter `guide`, the remaining frames come back unchanged.

## Outputs

| Output Name | Description | Data Type |
| - | - | - |
| `IMAGE` | Deflickered frames, with the same size and frame count as `images`. | IMAGE |

> This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/VideoDeflicker/en.md)

***

**Source fingerprint (SHA-256):** `d9937a67b0ac33ef3b326017e7d7990474705917902f1097cfeb06d9f83322a1`


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.