Elthibert
Newbie
- Apr 6, 2024
- 13
- 18
It depdends on what you want to achieve.
If your goal is to have a generated image that "perfectely" follows a complex prompt, you have two choices :
- Ideogram
- Dall e 3
Ideogram is right now at least as performant as dall e 3 in term of prompt complexity understanding, it's free and fast to generate images.
Dall e 3 is free using Bing but limited daily and you can only generate square format images. Using GPT +, you can use dall e with many aspect ration, and it's more useful to generate idea / try variants etc using the power of chat gpt.
In term of art style, Midjourney is currentely the best overall in terms or realism, but prompt understanding is below ideogram/dall e.
Finally, Stable Diffusion is more complexe and need more tweeking but you can have outstanding results. I would use it only if you are an artist to have your own unique style, or if you want to use it for porn related/NSFW generation.
In term of censorship :
- Dall e is a nightmare (even normal woman in bikini are sometimes censored, no artist name, no celebrity name, no blood etc....)
- Ideogram and midjourney are same level, pretty versatile, but no hot/porn
- Stable diffusion is uncensored.
In term of prompt complexity and length understanding :
- Ideogram / Dall e are the best
- Midjourney since maj v6 it understands better and way longer prompt
- Stable diffusion still really bad understanding, but is more used to simple prompt using high quality models (sdxl (juggernaut etc --> Find the best models using civitai website)
In term of style/art/consistency :
- midjourney is the best right now with the new --sref (style consistency) and --cref (characer consistency).
- then dall e is good for pixar 3d style to keep the same style but your character will generally have different face shape etc...
- finally Stable diffusion using Ipadapter feature but it's more complexe.
For stable diffusion, you have platforms like tensortart that are plug and play to learn the basics of stable diffusion (definitely check tutorials to learn the basic with controlnet etc...
If you want to become a pro with stable diffusion, use the websie thinkdiffusion wich is a cloud based computer with stable diffusion latests features.
Finally, Upsacale your images using Magnific ai, wich is currently the most advanced image upscaler in the world hands down. (expensive though)
Also check the website krea ai wich has a free upscaler based on the same technology, but the quality/realism does not come close to the one from magnific ai.
FInally stay up to date about image generation cause it's growing extremely fast. For example, stable diffusion 3 will soon be release and will be better than dall e 3... So everthing moves fast
If your goal is to have a generated image that "perfectely" follows a complex prompt, you have two choices :
- Ideogram
- Dall e 3
Ideogram is right now at least as performant as dall e 3 in term of prompt complexity understanding, it's free and fast to generate images.
Dall e 3 is free using Bing but limited daily and you can only generate square format images. Using GPT +, you can use dall e with many aspect ration, and it's more useful to generate idea / try variants etc using the power of chat gpt.
In term of art style, Midjourney is currentely the best overall in terms or realism, but prompt understanding is below ideogram/dall e.
Finally, Stable Diffusion is more complexe and need more tweeking but you can have outstanding results. I would use it only if you are an artist to have your own unique style, or if you want to use it for porn related/NSFW generation.
In term of censorship :
- Dall e is a nightmare (even normal woman in bikini are sometimes censored, no artist name, no celebrity name, no blood etc....)
- Ideogram and midjourney are same level, pretty versatile, but no hot/porn
- Stable diffusion is uncensored.
In term of prompt complexity and length understanding :
- Ideogram / Dall e are the best
- Midjourney since maj v6 it understands better and way longer prompt
- Stable diffusion still really bad understanding, but is more used to simple prompt using high quality models (sdxl (juggernaut etc --> Find the best models using civitai website)
In term of style/art/consistency :
- midjourney is the best right now with the new --sref (style consistency) and --cref (characer consistency).
- then dall e is good for pixar 3d style to keep the same style but your character will generally have different face shape etc...
- finally Stable diffusion using Ipadapter feature but it's more complexe.
For stable diffusion, you have platforms like tensortart that are plug and play to learn the basics of stable diffusion (definitely check tutorials to learn the basic with controlnet etc...
If you want to become a pro with stable diffusion, use the websie thinkdiffusion wich is a cloud based computer with stable diffusion latests features.
Finally, Upsacale your images using Magnific ai, wich is currently the most advanced image upscaler in the world hands down. (expensive though)
Also check the website krea ai wich has a free upscaler based on the same technology, but the quality/realism does not come close to the one from magnific ai.
FInally stay up to date about image generation cause it's growing extremely fast. For example, stable diffusion 3 will soon be release and will be better than dall e 3... So everthing moves fast
Last edited:
