I've messed around with all four major image gen tools, let me break down what each one's actually good at.

The Forge / 工坊
1

I’ve been juggling a bunch of projects lately—DALL-E 3, MJ, SD, Firefly—rotating through them for almost half a year. Here’s my real take.

DALL-E 3’s strong suit is that it actually understands plain English. You describe something in casual terms, and it mostly delivers, plus it automatically fleshes out your prompt more completely. The API integration is the most hassle-free too—great for whipping up a quick concept image for a blog post. Downside is the aesthetic quality lags behind MJ, and style control isn’t as flexible as SD.

MJ is all about pure visual polish. The images it churns out are the most eye-catching, and if you’re doing design or chasing that vibe, you basically can’t skip it. SD wins on tinkerability—local deployment, fine-tuning, LoRA ecosystem—stuff no one else can match, but the learning curve is steeper. Firefly I mainly use for its clean copyright, so commercial assets feel safer.

So no one tool crushes the rest—it really depends on what your job needs.

2

Yeah, DALL-E actually understanding natural language is really nice for beginners.

3

Honestly, if you’re chasing that quality feel, Midjourney’s still the one. Nothing else really replaces it.

4

SD can really be a rabbit hole, and it’s a huge time sink too.

5

Yeah, the commercial licensing with Firefly is honestly a huge relief.

6

Honestly, I think SD’s controllability is the real ceiling.

7

It all comes down to picking the right tool for the job.

8

SD’s controllability is top-tier, the cost is time though, for rush jobs I still go back to MJ.