📄️ Stable Diffusion
Generate and control brand-quality images with Stable Diffusion: SDXL, prompts, ControlNet, LoRA, inpainting, upscaling. 6 hours, a 40-question exam and a verifiable certificate.
📄️ 1. Latent diffusion components
Module 1 of the Stable Diffusion premium course: the three pieces of a latent diffusion model, the text encoder, latent sizes, what SDXL changes and a first generation with diffusers.
📄️ 2. Positive and negative prompts
Module 2 of the Stable Diffusion premium course: structuring a visual prompt, weighting terms, the honest role of a negative prompt, and the limits of text understanding.
📄️ 3. Steps, guidance, seed
Module 3 of the Stable Diffusion premium course: what each sampling parameter changes, Euler and DPM++ schedulers, reproducibility via seed, and the quality-versus-time trade-off.
📄️ 4. Text-to-image and image-to-image
Module 4 of the Stable Diffusion premium course: how image-to-image works, denoising strength as a dial, color variants of the reference chair, and chaining both modes.
📄️ 5. Inpainting and outpainting
Module 5 of the Stable Diffusion premium course: masked retouch on a local detail, canvas extension for a banner format, dedicated inpaint models and edge coherence.
📄️ 6. ControlNet
Module 6 of the Stable Diffusion premium course: structural conditioning with Canny, depth and OpenPose preprocessors, control weight, start and end steps, combining several controls.
📄️ 7. LoRA and textual inversion
Module 7 of the Stable Diffusion premium course: dataset of 20 to 40 images, LoRA training, trigger word, textual inversion by comparison, and detecting visual overfitting.
📄️ 8. Upscaling and face restoration
Module 8 of the Stable Diffusion premium course: diffusion upscalers versus dedicated networks, artifacts to watch, face restoration with CodeFormer or GFPGAN and print preparation.
📄️ 9. Compute cost and memory
Module 9 of the Stable Diffusion premium course: half precision, attention slicing, VAE tiling, CPU offload, memory versus resolution and wall-clock times on modest GPUs.
📄️ 10. Rights, provenance, acceptable use
Module 10 of the Stable Diffusion premium course: model and LoRA licenses, brand and personality rights, watermarking, C2PA metadata and an acceptable-use policy.
📄️ Recap and exam
Complete recap of the Stable Diffusion premium course: latent diffusion, prompts, ControlNet, LoRA, inpainting, upscaling, then the 40-question exam and the verifiable certificate.