There is a specific and widely shared misery in writing a good explanation and then needing a diagram to go with it. The thinking is done, the words are right, and now you are dragging boxes around in a tool that was not built for this. Napkin addresses exactly that moment. You paste your text, it proposes visual representations of what you wrote, you pick one and adjust it.
The scope is narrow and the execution is good, and the learning material is appropriately brief because there is not much to learn. The core interaction genuinely works. The suggested diagrams are usually reasonable interpretations of the text rather than decorative nonsense, and when the first suggestion is wrong the alternatives are one click away. For process descriptions, comparisons, timelines and simple hierarchies the hit rate is high enough that it saves real time.
That is a modest claim and it is true, which puts it ahead of a lot of tools making larger claims. Editability is the design decision that matters most and the one competitors get wrong. Output is a real object with movable elements, editable text and adjustable styling, not a rendered image you have to accept or regenerate. The tutorials cover this properly and it is the difference between a tool you use for finished work and a toy you use once.
Anyone who has tried to fix a typo in a generated image will understand why this is the deciding feature. The house style is the main aesthetic cost. Everything produced here is recognisable as having come from here, which is fine for internal documents and increasingly noticeable in published material as more people use it. Brand controls exist and do not go far enough to disguise the origin.
If your output is client facing and your brand matters, that is a real limitation, and the learning material does not raise it. Complexity is the functional ceiling. Simple structures work well and detailed technical diagrams do not, and the boundary arrives sooner than you expect. A system architecture with a dozen components, conditional flows and annotations is beyond it, and attempting one produces something that is technically a diagram and does not communicate.
Knowing to reach for a proper diagramming tool at that point saves frustration, and nothing in the documentation tells you where the line is. The thinness of the learning material is worth reading two ways. Positively, the product needs almost no instruction and the help centre is proportionate. Less positively, there is no material on getting more out of it as you improve, because there is not much more to get.
This is a tool you master in an afternoon and then use unchanged for years, which is a legitimate design goal and does mean the ceiling arrives quickly. The free tier is genuinely enough to evaluate it. Enough generations to test it against your actual work rather than a demo, which is the right way to decide, and more than several comparable tools allow. My advice is to run it against a document you have already written and struggled to illustrate, because that comparison answers the question immediately.
Three point nine. A focused tool that solves a small annoying problem properly, with learning material sized correctly for what it teaches. Marked down for a visual signature you cannot escape and a low complexity ceiling, and it is one of the few tools in this space where I think the product is honest about being small.