Diffusion Settings Explained: Steps, Seeds, Guidance
Understand the sliders in image tools so you change settings on purpose
A taste of a lesson
I set guidance to the maximum because I want it to follow my prompt exactly. Now the colours look crazy. Why?
Guidance pushes the image toward your prompt at every step, and at very high values it overshoots: colours saturate, contrast gets harsh and odd artefacts appear. Following the prompt more strongly is not the same as following it more accurately. Try this: keep the seed fixed and run three images at low, middle and high guidance. You will usually find the middle looks best and still matches your words. If the prompt is still ignored, the fix is usually clearer wording. Which of the three looks most natural to you?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Explain seed, steps, guidance and sampler in plain words
- Recognise signs of too few steps or too much guidance
- Generate at native resolution and know why it matters
- Run fair one setting comparisons with a fixed seed
Lesson plan
- 1 Seed: the starting point Use seeds to repeat results and to compare fairly. Start
- 2 Steps: how long it refines Choose a sensible step count for your model. Start
- 3 Guidance: how closely it follows you Set guidance so images follow the prompt without looking harsh. Start
- 4 Sampler, size and aspect ratio Understand samplers and why native resolution matters. Start
- 5 Negative prompts and strength Use negative prompts and image to image strength where available. Start
Try asking
About this tutor
For beginners who have opened the advanced settings of an image tool and seen steps, seed, guidance, sampler, aspect ratio and negative prompt with no idea what they do. This tutor explains each one in plain words, shows what happens when you push it too far, and gives you safe starting points and simple experiments. Not every tool exposes every setting, and names differ, so you learn the underlying ideas that let you recognise a setting wherever it appears. By the end you can make deliberate comparisons instead of random clicks.
Reviews
4.3
4 ratingsSample
- Babajide O.Sample
The two heads at high resolution explanation solved a mystery for me. Native size then upscale.
- Henrik A.Sample
Good but my tool hides most of these settings, so a lot of it was theory for me.
- Claire T.Sample
Finally understand what the seed does. Fixing it and changing one thing at a time made everything less random.
- Lucía M.Sample
Simple and patient. It mapped my tool's odd labels to the general ideas, which was exactly what I needed.
About the teacher
The technical side of image AI: local models, node pipelines, adapters and settings
9 tutors 331 lessons taught Sample
I teach the engineering side of image generation to people who want control rather than a single text box. I started as a hobbyist running open models on my own machine and later built image pipelines for small studios, so I know where the frustrations are: memory errors, inconsistent batches, settings nobody explains. I teach from first principles, then from...
See Quentin's profile and tutorsMore like this
Other tutors on the same or nearby topics.