Do I need a diffusion model or checkpoint for this workflow?

No checkpoint is required for grading—the GLSLShader applies your CurveEditor settings directly. For video, you do need the ByteDance2ReferenceNode (Seedance 2.0) and its dependencies. Without it, you can still grade and export the still image.

Is the GeminiNode required?

No. GeminiNode is optional and only helps draft or refine a motion prompt. You can type a prompt manually and send it to ByteDance2ReferenceNode. If you do use Gemini, make sure your API credentials are configured per the node’s instructions.

How can I avoid banding or color clipping when using curves?

Keep curves smooth with as few anchor points as needed, avoid extreme S-curves, and blend the grade with the PrimitiveFloat controls feeding the GLSLShader. Use ImageHistogram to keep blacks and whites within range, and export with a high bitrate in SaveVideo to minimize compression artifacts.

My video colors drift from the graded still. How do I keep the look consistent?

Confirm that the ByteDance2ReferenceNode uses the graded image output as its reference input. Reduce motion strength or stylization settings if available, keep a fixed seed for repeatability, and ensure no additional color transforms are applied downstream before SaveVideo.

Color Curves to Video - ComfyUI Workflow

This workflow builds cinematic color grades with precision and then animates your graded still into a short video. It starts with LoadImage and ImageHistogram in the Input And Histogram group to help you read the tonal distribution of your reference image. You then shape the look using multiple CurveEditor nodes in the Native Curve Editors group (master plus red, green, and blue) and send those curves to a GLSLShader in the GLSL Curve/LUT Grade group. The shader applies your per-channel remaps in real time, so you can iterate quickly while checking before/after with ImageCompare and live previews with PreviewAny. SaveImage lets you export the graded still if that’s all you need.

To create motion from the look you’ve crafted, the Motion Prompt And Seedance group uses a ByteDance2ReferenceNode (Seedance 2.0) that generates a video conditioned on your graded reference image and a motion prompt. A GeminiNode can optionally draft or refine motion prompts for you, while PrimitiveFloat nodes drive key parameters such as curve blend or shader strength for fine control. Finally, SaveVideo assembles the generated frames into a video file. The result is a streamlined grade-to-video pipeline that’s easy to tune and iterate.

FAQ

Sık Sorulan Sorular

Tüm iş akışlarını gör