Resources
The tools, prompts, and terms — curated, honest, and dated. Everything here is current as of 2026 and will move.
AI tools worth knowing
MidjourneyHosted
Painterly, art-directed images with a strong house look.
v7 era, 2026GPT ImageHosted
Prompt-following and legible in-image text.
2026Adobe FireflyHosted
Commercially safe generation inside the Adobe stack.
2026IdeogramHosted
Typography and logo-style text that stays readable.
2026RecraftHosted
Vectors, icons, and brand-consistent sets.
2026FLUX.1Local / Hosted
State-of-the-art open weights, strong prompt adherence.
dev + schnellComfyUILocal
Node-graph control over the whole pipeline.
power usersCivitaiModel hub
Finding checkpoints, LoRAs, and reference workflows.
check licencesPrompt templates
The six-slot brief
[subject], [action/pose], [setting], [lighting], [style/medium], [camera/lens] — fill each slot once, in that order. Steer with structure before you reach for adjectives.Product on seamless
studio product photograph of [object], on a seamless [colour] background, soft key light with a subtle violet rim light, shallow depth of field, high-end commercial lookOn-model character
[character sheet reference via IP-Adapter], [new pose], consistent design, same colours and proportions, clean studio lighting, neutral backdropGlossary
- Diffusion
- The process a model uses to turn random noise into an image over many denoising steps.
- Latent space
- The compressed mathematical space where the model actually works, before decoding to pixels.
- CFG scale
- How strictly the model follows your prompt versus improvising. Higher is more literal, often less natural.
- Denoise strength
- In img2img, how far the model is allowed to move away from your input image.
- ControlNet
- A conditioning model that steers composition from a guide — depth, pose, edges — beyond the prompt.
- IP-Adapter
- A method for using a reference image to carry style or identity into new generations.
- LoRA
- A small add-on that teaches a base model a specific style, character, or subject.
- Checkpoint
- A full base model file. The foundation everything else plugs into.
- Upscaler
- A pass that raises resolution; tiled upscalers add detail without hallucinating a new image.
- Seed
- The starting random number. Same seed and settings reproduce the same image.
Put these to work in the process breakdowns, and get the full method in the book.