Vocabulary
Model, interface, prompt, reference, seed, mask, inpainting, upscale, workflow and multimodality.
Private course · 6 hours · Introductory level
Generative image artificial intelligence
A new visual culture
AI replaces neither the eye nor project culture. It becomes an instrument for exploration, representation and transformation.
This course starts from the foundations, with no prior vocabulary required. It explains how an image is generated, how to choose a tool for the task and how to write an instruction that protects what matters: viewpoint, geometry, material, light and intention.
The course draws on real-world interior design scenarios: transforming an existing render, correcting a specific area, enlarging an image, composing a presentation, or animating a fixed-camera plan.
“The tool accelerates the image. The eye retains the direction.” Course principle
01Programme
Six hours to understand, experiment and build a method.
Understanding generated images
Probability, diffusion, latent space, the role of randomness and the limits of plausibility.
Writing a prompt
Turn a request into a specification: task, source, invariants, modifications, appearance, exclusions and output.
Choosing tools
Understand the respective functions of generators, editors, specialist engines and finishing tools.
Animate and orchestrate
Kling AI, Lumion and multimodal interfaces: from a fixed shot to a visual workflow.
Build your method
Final exercise, control checklist and organisation of a repeatable chain.
02Foundations
Name things before choosing buttons.
Model
The trained mathematical engine that produces or transforms the image.
Prompt
A structured instruction. For a professional, a mini specification.
Reference
The image that guides structure, material, style, object or composition.
Inpainting
The targeted transformation of an area selected by a mask.
Upscale
The enlargement of an image, sometimes with creative reconstruction of details.
Workflow
An ordered sequence of inputs, processes, controls and outputs.
03Tools
The right instrument for each function.
01Midjourney
Explore an artistic direction
A particularly effective tool for opening up possibilities, building an atmosphere, creating a moodboard and searching for a visual language.
- Concepts, atmospheres and variants
- Style References, Omni Reference et moodboards
- Use with caution when geometry must remain exact
02Nano Banana 2
Transform through conversation
A multimodal model suited to guided transformations, work with multiple references and successive corrections.
- Image-to-image and conversational editing
- Materials, objects, light and formats
- Invariants must always come before modifications
03Archsynth
Recalculate an architectural rendering
A specialist application for quickly transforming sketches and renderings into more realistic visualisations.
- Sketch-to-image, re-render and inpainting
- Material variants and moodboards
- Systematic checking against the source model
04Magnific
Enlarge and reconstruct detail
A finishing tool for upscaling, materials, relighting and style transfer.
- Faithful or creative upscale
- Stone, wood, textiles, metal and high-definition crops
- Check joints, veining, text and constructed details
05Kling AI
Give movement to the image
An image-to-video solution for short shots, setting a space in motion and time-based scenarios.
- Camera, trajectory, duration and continuity
- Fixed-camera animation
- Check mechanics, scale, shadows and object continuity
06Lumion
Visualise from the 3D model
A real-time architectural visualisation engine linked to the geometry, camera and materials of the 3D model.
- Images, animations and panoramas
- Continuity between views of the same project
- Distinct from a general-purpose text-to-image generator
07ChatGPT Images
Design and edit within the same dialogue
A conversational environment that can formalise the brief, produce the image, analyse variants and prepare deliverables.
- Guided generation and transformation
- Successive retouching and derivative formats
- Methodical comparison between source and result
08Claude Design
Build a visual system
A conversational design space for exploring visual experiences, commenting on and refining design directions.
- Presentations, interfaces and narrative boards
- Iteration through conversation and comments
- Distinct from a photorealistic architecture engine
04Interfaces
Assemble tools into a clear chain.
Orchestrate
Connect multiple models and operations within a visual node-based canvas.
Define the authority
Plan, rendering, photograph, or simple mood reference.
Modify
Change only what has been decided, while protecting the invariants.
Compare
Examine geometry, lighting, materials, and details at 100%..
Adapt
Portfolio, client presentation, vertical story, print, or video.
05Practical work
Exercises drawn from real situations.
From rendering to photography
Transform an old rendering into a natural image without changing its viewpoint or composition.
Local modification
Replace a floor, object or material while protecting everything outside the mask.
The moving image
Write a ten-second scenario with a fixed camera, trajectory and physical constraints.
The complete workflow
Organise the chain: source, transformation, retouching, upscale, layout and distribution.
06Information
A day dedicated to the eye and the method.
Duration
6 hours, including breaks
Level
Introductory — no technical prerequisites
Format
Private course combining explanations, demonstrations and exercises
Equipment
Laptop, recent browser and a few personal images to transform
Course material
Vocabulary, prompt templates, control checklists and recommended workflow