ComfyUI Professional Course for VFX

Who this course is for

For compositors and visual effects artists who want a practical route into ComfyUI and more control over generated imagery. The course begins with installation and the core graph, then develops techniques you can combine for image and video tasks.

Begin by establishing a ComfyUI environment for the exercises; local and cloud approaches are discussed during the course.

Course at a glance

10 classes and one bonus lesson. Total recorded time is 17:57:52, approximately 18 hours. Follow the classes in numerical order and watch the bonus between Classes 05 and 06.

What you will learn

• Build and inspect a ComfyUI workflow from its inputs through its saved output.
• Make deliberate choices about reconstruction, structural guidance, masking, and reference images.
• Integrate image descriptions and language-model outputs into a practical graph.
• Organize environments and dependencies so work can continue across machines.
• Understand the logic of video-generation workflows and document the context of AI-assisted work.

Class 01 Installing ComfyUI and understanding its environment

Recorded time 01:16:15

Start with a ComfyUI setup you understand. Compare installation approaches and the software and hardware components behind a working environment. Learn how dependencies, compatibility, and portable installations affect the setup you will use throughout the course.

• Distinguish desktop, cloud, and API modes of use.
• Recognize the roles of Python, PyTorch, CUDA, and supporting components.
• Compare installation approaches and identify missing workflow dependencies.

Suggested practice Set up an environment you can use throughout the course. Record its installation type and identify where to inspect dependencies.

Class 02 Core node graphs and image generation

Recorded time 01:22:51

Take control of your first image-generation graph. Explore the interface and follow the connections between the checkpoint, text conditioning, latent image, sampler, decoder, and saved output. Understanding this path helps you decide which setting to change and why.

• Read nodes, connections, and data types in a workflow.
• Trace a basic text-to-image generation from inputs to output.
• Run a workflow and investigate how its settings affect a result.

Suggested practice Rebuild the supplied cat-generation workflow. Change one setting at a time and retain comparable outputs.

Class 03 Configuring and organizing ComfyUI

Recorded time 01:52:31

Make your workflows easier to reopen, inspect, and develop. Organize folders and models, explore configuration and workflow utilities, and refine the basic graph so that inputs, outputs, and supporting files are easier to manage.

• Locate input images, generated outputs, and model files.
• Apply clear naming and organization to files and workflows.
• Use configuration and workflow utilities to support repeatable work.

Suggested practice Organize the example workflow and its assets. Create a clear output naming scheme and confirm you can reopen and continue the work.

Class 04 Comparing and controlling upscaling

Recorded time 01:55:01

Learn to judge what an upscale changes in an image. Compare deliberately reduced photographs with higher-resolution references, then explore model-based upscaling, tiled diffusion, sequential tile processing, and API alternatives. Use those comparisons to assess reconstruction and invented detail.

• Compare an upscale against a known source image.
• Distinguish reconstruction and invented detail when judging results.
• Explore different upscaling workflows and their controls.

Suggested practice Upscale one supplied low-resolution image with two approaches. Compare both with the higher-resolution source and note the differences.

Class 05 Structural control with ControlNet

Recorded time 02:05:03

Guide image structure more deliberately with ControlNet. Work through Canny, depth, and pose examples, understand the components and their compatibility, and explore how multiple controls work together. See how adjusting control influence changes the relationship between the source and the result.

• Distinguish the base model, control model, preprocessor, and control application.
• Build and inspect Canny and depth control workflows.
• Adjust control influence and explore combined controls.

Suggested practice Build a single-control workflow, vary its influence, and then compare it with a combined-control example.

Bonus Class Workflow auditing and production context

Recorded time 00:32:03

Make your workflow easier to review in a production context. Explore a workflow-audit system and a production-context declaration, connecting the practical graph with transparency and responsible-use principles. Watch this lesson after Class 05.

• Explain why a workflow needs production context and provenance information.
• Explore the supplied audit guide and declaration template.
• Review an audit critically and identify information that needs verification.

Suggested practice Use the supplied materials to document the context of a sample workflow. Identify missing information and questions that need further investigation.

Class 06 Language models and image understanding

Recorded time 02:19:33

Turn image descriptions into useful inputs for a graph. Build from primitive values and text handling to Florence2, Qwen VL, and an OpenAI API workflow. Explore image descriptions and prompt generation, while comparing the resource and data-handling implications of local models and APIs.

• Handle basic values and text within a node graph.
• Generate image descriptions and use them in a workflow.
• Compare local model and API approaches, including resource use and data handling.

Suggested practice Describe a supplied image with a visual-language model and use the description in an image-generation workflow. Compare the description with the source.

Class 07 Image to image and selective image editing

Recorded time 01:09:44

Direct changes to the part of an image that needs attention. Work with image-to-image generation, inpainting, outpainting, masks, alpha, segmentation, and background removal. Examine how model choice and mask transitions affect the way a localized edit fits the surrounding image.

• Distinguish full-image transformation from a masked edit.
• Prepare masks and use segmentation to select an area.
• Evaluate transitions and adapt the workflow to the specific image.

Suggested practice Change a selected region of a supplied image. Compare a manually prepared mask with a segmentation-based selection and inspect the transition.

Class 08 LoRAs and IP Adapters

Recorded time 01:21:55

Use model adaptation and image references to guide a generation. Compare LoRAs with IP-Adapters, see where each acts in the graph, and explore how their controls influence the result. Learn to inspect an example workflow before adapting it to your own intention.

• Distinguish LoRA model adaptation from IP-Adapter image conditioning.
• Explore their controls and their interaction with the base workflow.
• Investigate node and model documentation when adapting an example.

Suggested practice Create comparable results using a LoRA, an image reference through IP-Adapter, and a combination. Describe what each contributes.

Class 09 Moving workflows between computers and cloud environments

Recorded time 01:18:13

Plan how to keep working when the computer or environment changes. Identify the files and configuration a graph depends on, explore synchronization and shared model organization, and compare local and cloud options including RunPod, RunComfy, and Comfy Cloud.

• Identify the files and configuration needed to continue a workflow elsewhere.
• Understand synchronization and shared model-path organization.
• Compare local and cloud environments for a practical workflow.

Suggested practice Plan a transfer between two environments. List what needs to move or synchronize and verify that the workflow can locate its dependencies.

Class 10 Generative video concepts and workflows

Recorded time 02:44:44

Extend your understanding of the graph to moving images. Explore the logic and methods behind generative video, then examine ComfyUI examples, workflow variations, and subgraphs. Examples and references around LTX Video, Wan, and Seedance help you recognize recurring structures and decide what to test next.

• Identify different approaches to generating or transforming video.
• Trace the logic and components of a video-generation workflow.
• Recognize recurring workflow structures and explore subgraphs.

Suggested practice Choose a demonstrated video approach, produce a short test, and document its inputs, controls, and the changes you would investigate next.

Suggested capstone

Choose a small image or video task and build a workflow you can explain. Retain the input, the graph, the relevant settings, and comparable outputs. Describe your creative intention, identify the parts controlled by each component, and document limitations or questions raised by the result. Use the bonus materials to add production context.

The practice tasks and capstone above are suggested activities for this syllabus.

Begin with a working environment, then use each lesson to build a workflow you can explain and adapt. Bring your compositing judgement to the graph and make the creative choices yourself.