4 ms·

Midjourney has absolutely nothing to offer compared to proper finetunes. DALL-E has: it generalizes well (can make objects interact properly for example) and ha

Midjourney has absolutely nothing to offer compared to proper finetunes. DALL-E has: it generalizes well (can make objects interact properly for example) and has great prompt adherence. But it can also be unpredictable as hell because it rewrites the prompts. DALL-E's quality is meh - it has terrible artifacts on all pixel-sized details, hallucinations on small details, and limited resolution. Controlnets, finetuning/zero-shot reference transfer, and open tooling would have made a beast of a model of it, but they aren't available.

xionplean 3y ago
You obviously have no idea what you are talking about or just make stupid anime porn. For fine art MJ destroys everything and it is even close. I say this with using comfyUI/SD all the time. But you do you and your anime porn.
- orbital-decay 3y ago
  I'm actually a person making technical decisions (art decisions in the past) in a VFX/art studio, and I'm talking about production use. No generative AI currently passes any reasonable production quality bar, but is being tried by everyone for doing the work that can't be done or is cost-prohibitive otherwise, for example animation, long series with style transfer, filler assets creation etc. Anything that only has a text prompt can be discarded instantly. You have to be able to finetune it on your own material for consistency (of course I'm not talking about dubious 3rd party models), you need higher order guidance (e.g. controlnets, especially custom ones) and many other things. In the hands of a skilled person, a trivial Krita/Photoshop plugin (Firefly, SD, SD realtime) blows anything MJ can offer out of the water, simply because it has all that and you can't do much with text, it doesn't have enough semantic capacity to express artistic intent. I'm not even starting on animation. In fact, anything that involves non-explicitly guided one-shot generation of anything with light/shadow/colors/perspective is entirely out of the question with the current crop, because all models are hallucinating hard and aren't controllable within a single generation. There are attempts at fixing the perspective without explicit guidance, but it's going to be a long way and it's not super relevant to how things are done anyway. And for fine art, nothing beats a human painter, doing it by throwing prompts at AI mostly misses the point. I'm not even sure what you mean by fine art in this context, actually - surely not generating artsy-looking images from a prompt for fun?

xionplean 3y ago

You obviously have no idea what you are talking about or just make stupid anime porn. For fine art MJ destroys everything and it is even close. I say this with using comfyUI/SD all the time. But you do you and your anime porn.

orbital-decay 3y ago
I'm actually a person making technical decisions (art decisions in the past) in a VFX/art studio, and I'm talking about production use. No generative AI currently passes any reasonable production quality bar, but is being tried by everyone for doing the work that can't be done or is cost-prohibitive otherwise, for example animation, long series with style transfer, filler assets creation etc. Anything that only has a text prompt can be discarded instantly. You have to be able to finetune it on your own material for consistency (of course I'm not talking about dubious 3rd party models), you need higher order guidance (e.g. controlnets, especially custom ones) and many other things. In the hands of a skilled person, a trivial Krita/Photoshop plugin (Firefly, SD, SD realtime) blows anything MJ can offer out of the water, simply because it has all that and you can't do much with text, it doesn't have enough semantic capacity to express artistic intent. I'm not even starting on animation. In fact, anything that involves non-explicitly guided one-shot generation of anything with light/shadow/colors/perspective is entirely out of the question with the current crop, because all models are hallucinating hard and aren't controllable within a single generation. There are attempts at fixing the perspective without explicit guidance, but it's going to be a long way and it's not super relevant to how things are done anyway. And for fine art, nothing beats a human painter, doing it by throwing prompts at AI mostly misses the point. I'm not even sure what you mean by fine art in this context, actually - surely not generating artsy-looking images from a prompt for fun?