Thank You!!! the part of this video, was the Whole Damn Video..👍 i've been struggling to remove rocks from a tropical beach scene, and i'm sure this'll help.. 😃
well done. you don't mince words and give sufficient examples. you also provided me with some additional knowledge beyond just the --no command that i wasn't expecting and that was much appreciated. thanks yo
A thousand thank you's for this. I think one of my greatest challenges so far with AI image generators have been the negative prompts and writing them in such a way as to get the result I am wanting/seeking, especially in regards to head and feet getting cut off / out of frame quite a bit. You have given me some food for thought, things to try and re-word, and I thank you for that!!
Thanks! You're not the only one struggling with negative prompting. We all beat our heads against the wall sometimes trying to get the AI to listen. It's a bit like training a cat.
These are still “Wild West” days! But this is the fun part; being on the edge. It was the same when the internet first became viable. Now we take it all for granted!
My quick Midjourney v4 prompt guide: ➊ Use `::` followed by an optional weight to split the prompt into paragraphs. The order of the paragraphs doesn’t matter but their weights do: `office worker::` = `office worker::1`. Note: `--no glasses` = `glasses::-0.5`. The higher the weight, the more important is the paragraph. The lower (including negative), the less important it is. ➋ Within a paragraph, simple punctuation splits text into phrases: `blue office worker` ≠ `blue office,worker` ≠ `blue,office worker`. Repeated or different punctuation like `blue,,,office worker` or `blue.office worker` with the same seed produce small variation (because they technically constitute a slightly different prompt). Repeated spaces are the same as a single space. ➌ Within a paragraph, word order influences grammar & leftmost words or phrases matter more: `worker,blue office` ≠ `blue office,worker`.
You can get very funny results, for example "house::6 tree::1 door::-0.5 windows::-2 human, face::-3" will make MJ try to make a house without a door, which is tough for it, so in some images it'll try to COVER the door with something, and that something will be a human that looks a lot like Doctor House. Because that's still "house"
You must be reading my mind, Adam! I was just playing with this yesterday. There are some funny tricks with negative weighting. I'll try a few of your examples!
Have you had any success controlling limbs through negative prompts. Like if it's a shot of a model in a studio with her arm up, use negative prompts to move her arm down. Or if in image comes back with a person holding flowers, (or something), use negative prompts to say something like, 'nothing in the hands'.
You can try negative prompts. Also try adding a positive prompt. --no flowers in hand but also add "empty hands" in the prompt. This doubles up on what you want.
I've been trying to prompt --no lamp for kitchens and MJ is not having it. Kitchens got lamps. I am trying to create backgrounds for a lamp I already created so I can add them in Canva but MJ is like nope, there must be lamps.
In my experience, Midjourney NEVER pays attention to negative prompts. If you put "--no text", MidJourney will add text 95% of the time. (More often than if you don't add the negative prompt.) If you add "--no deformed fingers", MidJourney will almost always give you deformed fingers. If you want to use Midjourney images, it has been my experience that you have to generate at least 10 images to get one that is usable. And MidJourney's refusal to follow negative prompts is one of the main reasons.
Yes. I've had this experience, too. Did you see my suggestion about using BOTH a negative and positive prompt? Also, I have a new video on weighting (multi-prompts)
hi! thanks for this, what if i want the living or real life animals or creatures that i want? and not doll or puppets in my image? how can i achieve that with negative prompt?
See my video on the version 5 update. Everything is so much more real!! Make sure you're asking Midjourney for a photo. You can use photographic terms and equipment - like Nikon.
Yes. These are common on Stable Diffusion. On Midjourney, you don't need so many negative prompts. Mostly the MJ style takes care of the worst of the distortions.
Yes. This video is using the current version 4c. Image weights (--iw) don't work in version 4. Maybe that's what you are thinking of. Multiprompts (::) work.
On openartai does it make a difference if you publish the perfect creations so it will give the algorithm a better idea of what you do want to see and what you don't want to see?
It doesn't look at your images this way. The AI doesn't train in the images you create. Though there is some talk about getting the AI to do this for more personalized styles.
Are Midjourney images traceable, i.e. could someone find an image you made on the Midjourney platform just by searching? Is it easy to find? Note: I know that the images are not private (unless the user subscribes to a certain plan)
Your images are public on Midjourney. Their search function isn't that great, so it's difficult to find images that way. Openness is part of the culture of Midjourney. Sharing images and image prompts.
This is normal. Sometime the AI simply ignores the --no parameters. *Sigh*. The best way I've found to combat this is in the video. Use BOTH a negative prompt AND flip the idea over and phrase it in a positive way in the prompt itself. So, if you don't want text in the corner of your image, write "unsigned masterpiece" in the prompt and --no text, signature, words in the negative prompt.
Thank You!!! the part of this video, was the Whole Damn Video..👍 i've been struggling to remove rocks from a tropical beach scene, and i'm sure this'll help.. 😃
Thanks! 😃 Let me know how it goes removing the rocks!
Everyone's waiting for an update on the rocks
Very succinct and well-explained. Never could quite get my head around the whole double :: parameter until I watched this. Thank you!
Great! Keep creating, Bela!
well done. you don't mince words and give sufficient examples. you also provided me with some additional knowledge beyond just the --no command that i wasn't expecting and that was much appreciated. thanks yo
Great! Glad to hear it. I'm always looking for a new way to describe or use a tool.
A thousand thank you's for this. I think one of my greatest challenges so far with AI image generators have been the negative prompts and writing them in such a way as to get the result I am wanting/seeking, especially in regards to head and feet getting cut off / out of frame quite a bit. You have given me some food for thought, things to try and re-word, and I thank you for that!!
Thanks! You're not the only one struggling with negative prompting. We all beat our heads against the wall sometimes trying to get the AI to listen. It's a bit like training a cat.
Great video Jen. Very useful to know about the negative weight -1. Thank you!
Negative weights are great - when they work! I have a better result when I use both a negative and positive prompt to reinforce
These are still “Wild West” days! But this is the fun part; being on the edge. It was the same when the internet first became viable. Now we take it all for granted!
Yes! It does feel like that!
Great video☺️
May I ask what software you used to create your video. It's fantasitc.
I use Canva for most of my videos
It would be good to hear from you with the news on bluewillow v3....please
Thanks for the reminder!
Thanks for another great video! 👏
Thanks again!
Very informative thank you
My quick Midjourney v4 prompt guide:
➊ Use `::` followed by an optional weight to split the prompt into paragraphs. The order of the paragraphs doesn’t matter but their weights do: `office worker::` = `office worker::1`. Note: `--no glasses` = `glasses::-0.5`. The higher the weight, the more important is the paragraph. The lower (including negative), the less important it is.
➋ Within a paragraph, simple punctuation splits text into phrases: `blue office worker` ≠ `blue office,worker` ≠ `blue,office worker`. Repeated or different punctuation like `blue,,,office worker` or `blue.office worker` with the same seed produce small variation (because they technically constitute a slightly different prompt). Repeated spaces are the same as a single space.
➌ Within a paragraph, word order influences grammar & leftmost words or phrases matter more: `worker,blue office` ≠ `blue office,worker`.
It's important to understand: the prompt "fish --no fish" is the same as "fish::1 fish::-0.5". So effectively you get "fish:0.5" so it's still "fish"
"--no xx" is just a shorthand for "xx::-0.5". You can do multiple negative prompts with different weights, but the total sum of weights must be >0
You can get very funny results, for example "house::6 tree::1 door::-0.5 windows::-2 human, face::-3" will make MJ try to make a house without a door, which is tough for it, so in some images it'll try to COVER the door with something, and that something will be a human that looks a lot like Doctor House. Because that's still "house"
You must be reading my mind, Adam! I was just playing with this yesterday. There are some funny tricks with negative weighting. I'll try a few of your examples!
I'll try this. I haven't thought to push MJ in this way!
Great tips!
Thanks! Hope this helps!
Wow brilliant ! Thanks a lot ;)
Glad you like it, Lionel!
Have you had any success controlling limbs through negative prompts. Like if it's a shot of a model in a studio with her arm up, use negative prompts to move her arm down. Or if in image comes back with a person holding flowers, (or something), use negative prompts to say something like, 'nothing in the hands'.
You can try negative prompts. Also try adding a positive prompt. --no flowers in hand but also add "empty hands" in the prompt. This doubles up on what you want.
@@MakingPhoto Thanks
I've been trying to prompt --no lamp for kitchens and MJ is not having it. Kitchens got lamps. I am trying to create backgrounds for a lamp I already created so I can add them in Canva but MJ is like nope, there must be lamps.
Just Remarkable
Thank you, VISUAL NEXUS! Keep creating!
thank you!!
Just found out it yesterday 😀
😃We're in sync!
In my experience, Midjourney NEVER pays attention to negative prompts. If you put "--no text", MidJourney will add text 95% of the time. (More often than if you don't add the negative prompt.) If you add "--no deformed fingers", MidJourney will almost always give you deformed fingers. If you want to use Midjourney images, it has been my experience that you have to generate at least 10 images to get one that is usable. And MidJourney's refusal to follow negative prompts is one of the main reasons.
Yes. I've had this experience, too. Did you see my suggestion about using BOTH a negative and positive prompt? Also, I have a new video on weighting (multi-prompts)
@@MakingPhoto I'll look at the weighting video. Using negative weights didn't help, but I haven't tried BOTH yet. Thanks.
hi! thanks for this, what if i want the living or real life animals or creatures that i want? and not doll or puppets in my image? how can i achieve that with negative prompt?
See my video on the version 5 update. Everything is so much more real!! Make sure you're asking Midjourney for a photo. You can use photographic terms and equipment - like Nikon.
TQ!!
No problem!
if i uploaded a specific image from my computer how do i delete an object from it and keep image as it is - without depending on generative adobe ai
I used often:
2 heads, 2 faces, cropped image, out of frame, draft, deformed hands, signatures, twisted fingers, double image, long neck, malformed hands, multiple heads, extra limb, ugly, poorly drawn hands, missing limb, disfigured, cut-off, ugly, grain, low-res, Deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, floating limbs, disconnected limbs, long body, disgusting, poorly drawn, mutilated, mangled, surreal, extra fingers, duplicate artefacts, morbid, gross proportions, missing arms, mutated hands, mutilated hands, cloned face, malformed,duplicate, copy, multi, two, disfigured, kitsch, ugly, oversaturated, grain, low-res, Deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, missing limb, blurry, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, ugly, disgusting, poorly drawn, childish, mut
Yes. These are common on Stable Diffusion. On Midjourney, you don't need so many negative prompts. Mostly the MJ style takes care of the worst of the distortions.
oh oh That's good knowledge. I'll put it to good use.
--> 😊😊😊::2 😮😮😮::-2
😀😜
thank you for this , but does weight works in V4?
Yes. This video is using the current version 4c. Image weights (--iw) don't work in version 4. Maybe that's what you are thinking of. Multiprompts (::) work.
On openartai does it make a difference if you publish the perfect creations so it will give the algorithm a better idea of what you do want to see and what you don't want to see?
It doesn't look at your images this way. The AI doesn't train in the images you create. Though there is some talk about getting the AI to do this for more personalized styles.
It will happen eventually I'm pretty sure as it's come on really quickly in just the last Twelve months.@@jaade77
awesomw!
Thanks!
Are Midjourney images traceable, i.e. could someone find an image you made on the Midjourney platform just by searching? Is it easy to find? Note: I know that the images are not private (unless the user subscribes to a certain plan)
Your images are public on Midjourney. Their search function isn't that great, so it's difficult to find images that way. Openness is part of the culture of Midjourney. Sharing images and image prompts.
V5 working strange for me, --no not working sometimes and even works in oposite way.
This is normal. Sometime the AI simply ignores the --no parameters. *Sigh*. The best way I've found to combat this is in the video. Use BOTH a negative prompt AND flip the idea over and phrase it in a positive way in the prompt itself. So, if you don't want text in the corner of your image, write "unsigned masterpiece" in the prompt and --no text, signature, words in the negative prompt.
'Prompting disabilty' will make you want to smash your screen.
I'm a little afraid!!