icon icon

ComfyUI and RunComfy: AI Visual Lab – graphics and video with AI

icon

Training process

Training needs analysis

If you have specific requirements regarding the training programme, we will carry out a training needs analysis for you. This will guide us on which aspects of the programme should receive greater emphasis, so that the training programme meets your specific needs.

What will you gain?

icon

Smart AI tool selection - You will learn how to tell image, styling and video tools apart, so you can choose the right model or workflow faster for a campaign, design task or social media content piece.

icon

Confident work in ComfyUI - You will get comfortable with ComfyUI and RunComfy, understand nodes and data flow, and be able to launch, adjust and save workflows on your own instead of guessing each next step.

icon

Polished image generation - You will build text-to-image workflows, set models, prompts and generation parameters, and create visuals that match a brief, brand style and communication goal rather than random outputs.

icon

Precise creative control - You will learn to use ControlNet, LoRA and Refiner to guide composition, style and details more accurately, giving you results that are more predictable, repeatable and easier to refine.

icon

Effective image editing - You will work with image-to-image, inpainting and upscaling, so you can fix errors, restyle visuals, fill missing areas and raise the quality of finished graphics without starting from scratch.

icon

Animation from still images - You will explore image-to-video methods such as AnimateDiff and frame interpolation, allowing you to create short animations, loops and visual assets ready for online publishing.

icon

Automated content pipelines - You will connect nodes into larger pipelines, create reusable presets and render image batches, which helps you cut production time and scale recurring creative tasks much more efficiently.

icon

Responsible AI usage - You will understand licensing, copyright and deepfake-related risks, so you can use models and generated assets in a more responsible way and make better decisions in commercial work.

Training programme

1.

ComfyUI is introduced as a node-based environment for building image generation workflows with more control than prompt-only tools. The section defines key concepts such as nodes, checkpoints, prompts, samplers, latent images, and places the platform among tools like Midjourney, DALL·E, and RunComfy.

2.

A stable working setup starts with hardware and software requirements, then extends to installation choices and folder structure. The material also covers local versus cloud operation, using RunComfy, adding models and extensions, and resolving the technical issues that appear most often.

3.

Before more advanced workflows make sense, users need to navigate the graph-based workspace fluently. This section develops practical control of the ComfyUI interface: adding, connecting, and removing nodes, saving and loading workflows, organizing the canvas, and reviewing generated files.

4.

Creating the first image in ComfyUI requires linking several core parameters into one coherent generation path. The section covers base model choice, positive and negative prompts, resolution, sampler setup, steps, seed, CFG scale, and how these settings affect the initial output.

5.

Prompt quality directly influences composition, style, detail, and the frequency of unwanted artifacts. Participants work on writing effective prompts, using negative prompts, describing lighting and structure precisely, spotting common mistakes, and testing multiple prompt variants in a controlled way.

6.

Model selection shapes the visual character, technical behavior, and suitable use cases of the entire workflow. This part distinguishes Stable Diffusion families such as SD 1.5 and SDXL, maps models to output types, and covers installation, organization, and safe download practices.

7.

LoRA models are used to inject specific visual traits without changing the main checkpoint. The course explains when this approach is useful, how to install and load LoRA files, how to combine several of them in one workflow, and how to tune their strength for stylistic consistency.

8.

Instead of starting from an empty latent image, the process can begin with an existing graphic, sketch, or draft. This module explains style transformation, sketch-to-illustration workflows, denoise settings, and methods for balancing creative change with similarity to the source image.

9.

Editing selected parts of an existing visual is handled through masking and targeted regeneration. Participants learn to remove unwanted elements, add new objects, replace backgrounds, and correct faces, hands, or small details, with attention to practical mask preparation for controlled inpainting.

10.

When a layout needs more space, the image can be extended beyond its original frame without rebuilding it from scratch. The section focuses on outpainting, generating missing areas, adapting aspect ratios for banners or social media, and preserving consistency across the expanded composition.

11.

ControlNet adds structural control where prompting alone is not precise enough. The training covers pose guidance, contour and depth control, sketch-based generation, and maintaining object placement, showing how composition can be stabilized for commercial visual production.

12.

Higher-resolution output is not only about enlarging an image, but also about preserving clarity and usable detail. This section compares upscaling methods, improves sharpness, reduces artifacts, and prepares files correctly for online publishing as well as print-oriented delivery.

13.

Building a custom workflow means structuring the full path from prompt to final output around a specific use case. Participants design graph logic for marketing graphics, product visuals, and portraits, while also reducing unnecessary nodes and keeping workflows readable and documented.

14.

Additional functionality in ComfyUI depends on well-managed extensions and compatible dependencies. Here the work centers on ComfyUI Manager, installing and updating custom nodes, resolving missing-node issues, and applying practical rules that reduce instability when expanding the environment.

15.

To handle repetitive production efficiently, the program moves into batch-oriented image generation. It explains how seeds affect variation, how to build visually consistent series for campaigns, how to save reusable production settings, and how to shorten turnaround time without losing control.

16.

Generating short AI video sequences requires a different workflow than still image production. This part covers image-based motion, frame sequence handling, core video pipelines in ComfyUI, typical technical limits, and where such outputs fit in promotional and communication materials.

17.

RunComfy is presented as a cloud-based option for teams that want to avoid local installation and infrastructure setup. Participants work with ready-made templates, manage projects online, export results, and compare when cloud execution is more practical than a local ComfyUI environment.

18.

Business use is framed through concrete deliverables created with AI-assisted visual workflows. The focus includes social media graphics, ad creatives, product visualizations, moodboards, presentation materials, and illustrations for articles or websites, with attention to fit-for-purpose output.

19.

The section outlines the final operating guidelines for responsible AI image work in business settings. It addresses model licenses, rights to generated visuals, image and likeness risks, as well as file labeling, archiving, quality checks, compliance review, and ethical use of generative tools.

What are the prerequisites for participating in the training?

icon

Basic computer skills - You should be comfortable using your operating system, installing applications, managing files and folders, and working in a web browser without needing extra technical assistance.

icon

Basic graphics literacy - You should understand core digital image concepts such as file format, resolution, aspect ratio and quality, so you can assess generated results in a clear and practical way.

icon

Technical English - You should understand simple technical terms and interface messages in English, because tool interfaces, model names and many workflows rely on English labels and wording.

icon

Experimental mindset - You should be ready to test settings, compare variations and draw conclusions from results, because effective work with generative AI depends on iteration and patience.