Adobe's position in generative imagery is peculiar and worth understanding before you decide whether this training is for you, because the training only makes sense in light of it. Adobe was late and its models have generally not matched Midjourney or the best open weight models on raw output quality. What Adobe has instead is distribution and indemnification. Firefly is built into Photoshop, Illustrator and Express, where the working population of professional creatives already spends its days, and it is trained on Adobe Stock and licensed content, which lets Adobe make commercial safety claims its competitors cannot.
That is the strategy, and this training is an expression of it. The material comes through two channels. On Coursera there are Adobe courses including AI Tools for Creatives powered by Adobe Firefly and Prompting for Generative AI, both free to audit with an optional certificate fee. On Experience League, Adobe's own learning hub, there is a free AI training section with shorter hands-on modules.
The Coursera material is the more structured of the two and the Experience League content is better if you want to look one thing up. Production quality is good throughout, which is unsurprising given that this is Adobe and that its customers are people with visual taste. The prompting instruction is better than I expected and this is the part I would highlight for anyone dismissing this as marketing. Prompting for image generation is not the same skill as prompting a language model and the differences are not obvious.
Composition, lighting, lens and focal length language, style referencing, aspect ratio, negative space, the way structural terms behave differently from descriptive ones. A lot of people arrive at image models having got good at chatting with a text model and find their prompts producing nothing like what they pictured, because they are using the wrong vocabulary. The course treats this as a craft with a specific language, which it is, and much of that language transfers to other tools even though the specific interface does not. The workflow coverage is the other genuinely practical strength.
Firefly Boards, generative fill and expand, the integration points with Creative Cloud, how generative work moves between the applications in a real project. If you are a working designer, this is the part that changes your week, because the question is never can the model make an image, it is how does this fit into a file I have to hand to a client on Thursday. Product training is often derided and it is exactly what a professional using the product needs. Then the honest limitation, which is the same fact stated from the other side.
This is product training. A significant share of what you learn is where Adobe put the buttons. It does not transfer to Midjourney's parameter syntax, to Stable Diffusion checkpoints and LoRAs, or to ComfyUI's node graphs, and it teaches you very little about how any of it works underneath. You will not come away understanding diffusion, latent space, sampling steps or guidance scale, and if you want to actually understand generative imagery rather than operate one implementation of it, an open toolchain will teach you far more because it exposes the machinery.
The ethics and sourcing material needs to be read with one eye open. Adobe trained Firefly on Adobe Stock and licensed content, and it offers enterprise customers indemnification against copyright claims. That is a real and significant commercial differentiator, and for anyone producing client work or operating inside a company with a legal department, it is possibly the most important fact about Firefly. So I am glad the training covers it.
But the framing throughout is Adobe explaining why Adobe's approach is the responsible one, and that argument is simultaneously true and a sales pitch. The parts about provenance, content credentials and what you can and cannot safely use commercially are worth your attention. The implicit conclusion that this makes Firefly the correct choice for all work is Adobe's opinion, and a designer doing personal or exploratory work may reasonably weigh output quality more heavily. The quality gap is the thing the training will never mention and that you should know going in.
Firefly has improved considerably and its recent models are much more competitive, but on aesthetic ceiling it has generally trailed the leaders. For a lot of professional work this does not matter, because the job is a background extension, a clean object removal or a quick concept board, and Firefly integrated into Photoshop beats a better model that requires exporting and reimporting. For work where the image itself is the deliverable and needs to be striking, you may find yourself reaching elsewhere. Material dating is the last caveat and it is generic to the category.
Adobe ships new Firefly model versions and reorganises the product regularly, so screenshots and menu paths drift out of date. The prompting principles survive that. The click-here instructions do not. My 3.6 is a rating of well made, free, honest-about-its-scope product training.
It is exactly what it says it is and it does that competently, and the licensing material in particular is the kind of thing that saves a professional an actual problem rather than teaching them a trick. It loses points for being narrow by design, for teaching an interface rather than a discipline, and for a sourcing story that is both a genuine strength and a piece of positioning. If Adobe's tools are already your working environment, this is a straightforward yes at a price of nothing. If you are trying to learn generative imagery as a subject, start somewhere more open and come back to Firefly when you need the commercial safety.