Course Content
13 Lessons
Loading
This workshop focuses on mastering Flux models, ComfyUI, AI architectural visualization, multimodal workflows, and 3D design techniques.
4 courses5.0
Generate architectural images using the full Flux model suite in ComfyUI.
Build fast text-to-image workflows with Flux Schnell.
Apply LoRAs and image references for custom styles.
Guide designs with ControlNets and refine using flux fill.
Combine tools into advanced, flexible workflows.
Edit images with text and visuals using Kontext Dev.
Create animated architectural videos with Kling AI.
Turn 2D images into 3D models using Hunyuan 2.5.
Focus: Building a deep understanding of the core Flux models and mastering the essential techniques for image generation, control, and refinement in an architectural context.
Presentation: The Diffusion Architect 3.0 – Why the Flux Era?
Overview of the Flux architecture: What makes it a paradigm shift from previous models (speed, efficiency, transformer base).
Introduction to the specific Flux variants and their roles:
schnell (speed)
dev (quality/control)
fill (utility)
kontext dev (multimodality)
Text-to-Image with flux schnell
Creating a foundational text-to-image workflow in ComfyUI.
Exploring the speed and characteristics of flux schnell for rapid architectural ideation and concept development.
Prompting strategies tailored to schnell's behavior.
Tool Focus: ComfyUI, flux schnell.
Image-to-Image with flux schnell
Adapting the workflow for img2img tasks.
Understanding key parameters (denoise) for stylistic transfer and refinement.
Image-to-Image + ControlNets with flux dev
Introducing flux dev for higher-fidelity outputs.
Integrating ControlNets: Why flux dev is often preferred for controlled workflows.
Architectural Control Tools: flux canny + flux depth
Practical application of canny (from sketches/line drawings) and depth (from simple 3D massing) using flux dev workflows.
Tool Focus: ComfyUI, flux schnell, flux dev, ControlNet (canny, depth).
Deep Dive into flux loras
Understanding how LoRAs are implemented and interact specifically with Flux models.
Sourcing, applying, and weighting architectural LoRAs for targeted stylistic control (e.g., Brutalism, Biophilic design, specific material palettes).
Image Referencing with flux redux (IPAdapter)
Mastering image-to-image referencing workflows for consistent style, mood, and compositional guidance.
Balancing prompt, LoRA, and redux influences.
Tool Focus: ComfyUI, flux dev, Flux-compatible LoRAs, flux redux (IPAdapter).
Corrective Inpainting with flux fill
Introduction to flux fill for seamless inpainting and outpainting.
Architectural use cases: Refining details, removing unwanted artifacts, extending canvases, and targeted design modifications.
High-Resolution Upscaling with flux dev
Techniques for upscaling generated images for presentation while preserving detail and coherence, using workflows centered on flux dev.
Review of the core Flux workflows and techniques covered.
Open Q&A session.
Focus: Synergizing the learned techniques into complex workflows, exploring multimodal generation, and translating 2D masterpieces into dynamic video and 3D forms.
Building the Ultimate Architectural Workflow
This hour is dedicated to constructing a complex, synergistic workflow from the ground up.
Layering multiple techniques for maximum creative control:
flux dev as the base
Conditioned by multiple ControlNets (flux canny, etc.)
Guided by flux redux referencing
Stylized by flux loras
Strategies for organizing and managing complex ComfyUI graphs.
Live demonstration: Creating a highly detailed and specific architectural scene using the full combinational workflow.
Tool Focus: All Day 1 tools combined.
Exploring the Frontier: Multimodal Interaction
Deep dive into flux kontext dev: Understanding its unique ability to process both image and text prompts simultaneously for in-context editing and generation.
Demonstrating powerful use cases:
Style Transfer: "Make this photorealistic image look like an anime sketch."
Object/Subject Tweaks: "Change the building material from concrete to timber."
Text Editing: "Change the sign on the building to say 'Library'."
Comparing flux kontext dev to traditional inpainting and img2img methods.
Tool Focus: ComfyUI, flux kontext dev.
Bringing Architecture to Life
Taking high-quality images generated from the combinational or multimodal workflows and translating them into video.
Workflow and best practices for preparing images for Kling AI.
Exploring Kling's capabilities for creating dynamic camera movements, atmospheric effects, and short architectural fly-throughs.
Tool Focus: Kling AI; ComfyUI (for source image preparation).
Image-to-3D with Hunyuan 2.5
The final leap: Using refined 2D images as a source for 3D model generation.
Introduction to Hunyuan 2.5: Capabilities, ideal input characteristics, and workflow.
Step-by-step process of generating a 3D model from a selected architectural image.
Assessing & Integrating the 3D Output
Critical analysis of the generated 3D model's geometry, texture, and suitability for architectural conceptualization.
Brief overview of potential next steps (e.g., cleanup in Rhino/Blender).
Recap of the entire Flux ecosystem journey.
Discussion on the future trajectory of generative AI for architects.
Final Q&A session.
Tool Focus: Hunyuan 2.5
Course Content
Ismail Seleit is a distinguished architect at Foster + Partners, renowned for his expertise in computational design and BIM (Building Information Modeling). His work encompasses various projects, from design competitions to complex building realizations. Ismail is dedicated to integrating advanced AI tools, such as Stable Diffusion and ControlNet, into architectural workflows, pushing the boundaries of design innovation. Beyond architecture, he is an accomplished ambient-electronic music producer, creating soundtracks for architectural films and various independent projects, showcasing his multifaceted creativity and technical skills.
Very informative and helpfull even for someone who has some knowlege on the themes discussed
Short Update for Flux 2?
Good overview of the flux models and impact of LoRAs. Would be great if there is a separate workshop solely focused on trainig LoRAs.
Amazing workshop
Very practical workshop !
excellent course
Super good course
Great WORKSHOP
Nice course.
Great course! insightful and practical
nice tutors and workshops! but a little bit difficult website to get certificate...
amazing course, Ismail is a great instructor
The course was amazing, as a first timer in comfyUI and all these tools I found the way Ismail explained was very easy to understand and I was able to follow up with everything he has done. I will be watching everything twice maybe more to fully grasp everything but in general was mind blowing! Thanks a lot!!
Very informative session with Ismail, a great lecturer, liked how packed the sessions were, it was relatively easy to follow, although I think it needs to be clarified that attendees need to be familiar with comfyUI.
