Speaking honestly as someone who has used all three:
I think competition will expand into video generation down the line. Once Sora goes commercial, the landscape will completely shift.
The forced aspect ratio in Midjourney is really annoying lol, but I have to admit the quality is good.
Running Stable Diffusion locally makes the GPU fan noise absolutely insane. Still, there's a certain fun in creating custom models.
Well, rather than saying DALL-E 3 lacks creativity, isn't it actually more practical because it has fewer constraints? Midjourney does produce pretty images, but the weird fingers are still an issue.
Sora is going to be a real game changer when it comes out. But personally, I'm more excited about photorealistic 3D modeling than video generation.