@@TokenizedAIIt's actually a welcome and comforting/calming relief from the overly enthusiastic and energetic UA-camrs... some (or more) who are just faking it for the camera.
Great methodical approach with excellent examples explained perfectly! This is exactly this kind of info I've been looking for but couldn't find. Thank you!
I don't usually make comments but I thought it important to leave here my record and thanks for speaking so clearly. English is not my native language, but I understood every word you said, so here's my thanks and keep up the excellent work.
I tried it three days ago, I got really amazing results. Cinematic lighting. Backlighting. Natural lighting.... And lots of keywords to get what you want.
*i've only very recently(ish) started messing about with ai art generator programs and have only experimented with about three of them so far but i really like what the results have been...the randomness has been interesting and personally the distorted/disjointed surreal facial assemblages at times give the images an other worldly aspect that suits the artistic narrative i find appealing...i have found that the order in which i input my prompts also effects the final results as the first prompt is given more weight than the others...*
@@TokenizedAI *easy enough to test this theory...generate an image using a set of prompts then then switch the order around after a few images...maybe even have the same prompt more than once but at a different place such as at the beginning and then near the end*
Nicely done tutorial, thanks! How about a tutorial on widely used prompt words for styles. With or without: cyberpunk, psychedelic, necromancer, surreal, vaporwave, etc.
Great suggestion! But just to confirm. You want to know what all these difference styles are and what they look like? What specific "problem" should the video address?
Super helpful, thank you! I've just recently started using Midjourney to create art concepts for my film projects. I've been getting some great results that look very realistic and filmic, but I'm still struggling a lot with more detailed framing attempts - like camera angles, positioning of characters within the frame, foreground and background and so on. I'll keep a lookout on your channel to see if you ever discuss these topics. Thanks once again!
There are effectively two ways to achieve what you want to do. You either need a super elaborate prompt and some luck. Or you need to use multiprompts. But those are challenging and not easily explained in a comment section. I will be doing a video on this, rest assured.
Perhaps it could help starting with an image prompt. Take a picture reference roughly in the same angle, framing and composition you want and start generating variations with that as a base.
Just started playing around with this myself. I think commas do still have a place. For example you can add weights to entire sentences and/or single adjectives alike: A majestic black horse on top of a mountain::5, sunset::2, etc etc. In this example it isn't just giving "mountain" a weight of 5 but that entire sentence.
Yeah, but it's not doing that because of the commas. It's doing that because that how weights always work in multiprompts. The weight applies to the full sentence, so everything left of the colons....until it hits another double-colon. By the way, I use commas too. But primarily to makes things more easy to read. But they don't really do anything.
@@TokenizedAI It’s slightly confusing then, if every word has a default of ::1 (even if it’s not typed) and you give the last keyword a higher value it’s going to effectively make everything to the left of it that higher value. To me it would make sense that a comma could act as a roadblock to cancel out that effect. Because in massive prompts I don’t really wanna go through and assign everything a weight, just a handful of spots as needed to tweak & boost.
I think there a misunderstanding here. The examples I used only had singular keywords, but the "keywords" can be entire phrases as well. Example: /imagine a big house::1 spotless blue sky::2 a field of roses::3 The above example is split into 3 segments. - a big house (weight of 1) - spotless blue sky (weight of 2) - a field of roses (weight 3) Individual words don't have their own weight unless you turn them into a segment by spliting it with ::
@@Mimeniia That's incorrect. This is exactly how multiprompts work. Colons separate your prompt into segments and the weights apply to the entire segment, irrespective of whether it's only a single word or an entire phrase.
@@TokenizedAI But honestly this was actually helpful laid out like this as it reminded me just how important and useful they can be and I’ll use it more. I didn’t expect to get as much out of it as when I started the video. I enjoyed it to the extent that I checked I was subscribed! So 👍 ☺️
SUMMARY of the video by CHATGPT: The speaker is discussing the importance of prompt structure when using mid-journey text weights. They explain that many people believe that structuring prompts with commas separates them into different concepts, but that this is not true. They provide examples of prompts that are the same despite the use of commas and explain that mid-journey does not pay much attention to punctuation or sentence structure when interpreting prompts. They suggest that a basic rule of thumb is to not rely on commas to separate concepts in a prompt, and instead to use text weights to control the output. The speaker continues by explaining that text weights can be used to control the output of prompts by giving different weights to different keywords or concepts in a prompt. They provide several examples of prompts with weightings and compare them to the same prompts without weightings. They show that the use of weightings results in more creative and varied output. The speaker also explains that text weights can be used with full sentences and that the results can be surprisingly good. They use the example of a hacker fighting with a tree and show that the output is more creative and varied than without the use of weightings.
Midjourney is a godsend for workflows, productivity and creative output, thank you for helping us harness the possibilities even more "precisely" to our imaginings! : )
If you want a hacker fighting with a tree. You should also think about the sentence more. "a hacker fighting a tree" or "a hacker fighting against a tree" would get more consistent results (and less hackers with trees).
Can you add weights to reference image links? I’m trying create a cyberpunk character on horse but struggling to get midjourney to reference my character and horse images equally
I've found that for pics with lots of inputs and remixes that you have to really crank up the weight to get a palpable change. even up to the max of 10000. Usually I'll start with small numbers, especially when the prompt is still small. after many remixes it could be very big where I find I'll crank the numbers up as I remix it to keep pushing those aspects of the work that are still lacking. I've found too that when the prompt is very small, like say just 2 words, that it tends to overpower to one side or the other. if you add some benign additions and give them small weights too it can help to better balance the numbers in those earlier stages of crafting the prompt.
That's because the weight values aren't absolute numbers. They are relative to the overall number of weights you've added. Say you have 4 words with 4 different weights (2,1,4,3). The total weights are 10. So im that context, 3 is actually just 30%. If you have less words, then it will likely be stronger.
I haven't really used Blue Willow much and if you check Google Trends, the interest in it is nowhere near the same level. I think mainly because it's effectively the same thing. It also uses Stable Diffusion under the hood and copied most of Midjourney's workflow. So I'd assume that they'll also copy most of the parameters as well, which means that same prompts might work in a similar way.
Nice one my friend thank you. Does this weighting work with an image url too. ( I guess I should just go and try.. ) Thanks Again for your great videos
I love your calm, clear way of introducing this information.
I guess it's just the way I always talk 😅 I'm not particularly good at hyperactive engagement-inducing "on the edge of your seat" presentation 🤣
@@TokenizedAIIt's actually a welcome and comforting/calming relief from the overly enthusiastic and energetic UA-camrs... some (or more) who are just faking it for the camera.
Great job, keep uploading videos about midjorney, you are the best
Thanks so much!
I love that your tutorial consist so understandable infographic. It’s very convenient to watch it in phone, thank you
Thanks for this great feedback :)
I agree too bro nice job
Great methodical approach with excellent examples explained perfectly! This is exactly this kind of info I've been looking for but couldn't find. Thank you!
Glad it was helpful!
"and most importantly, CONTROL" sure sure, i see it clearly in that six finger fist
🤣 Some things are just out of our control.
😂
Dawg 😂😂😂
That extra finger makes all the difference.
Hrithik Roshan
Awesome and thorough explanation of a very useful concept! Thanks a lot for sharing your knowledge with us!
Glad it was helpful!
holy shit, this changes EVERYTHING. I can get exactly what I want much more quicker now
I'm not sure it changes everything. But it changes a lot :D
Extremely helpful. 10/10. Will watch again 😂
thank you so much this really completley changed my prompts technique and how i use it
Glad it helped! 😀
Wow. This was a huge help with Midjourney. Thank you!!
Glad it was helpful :)
Very useful. I like the big prompt graphics. Extra work true, but good.
I think it really helps people understand what's going on.
I really appreciate your Midjourney Videos, you explain all content, in way that I can easily understand!! Newbie here … Thank you 💕😊
You are so welcome!
Not even 1 min in and already learning, thank you!
Great to hear! 😁
Man, your videos are top notch! TY👍
I appreciate that!
man, you are really opens my eyes. thank you
Glad to hear it 🙂
the best midjourney tutorial ive seen! thank you
Wow, thanks! 🤗
Thank you for this. This is what I was looking for!!
Glad it was helpful!
I don't usually make comments but I thought it important to leave here my record and thanks for speaking so clearly. English is not my native language, but I understood every word you said, so here's my thanks and keep up the excellent work.
My pleasure! :)
Man, that's revolutionary, it changes everything.
It does, doesn't it? :)
You just upped my game. Thank you sir!
Awesome! 😀
I tried it three days ago, I got really amazing results.
Cinematic lighting. Backlighting. Natural lighting.... And lots of keywords to get what you want.
Yep, it's really great for creating more nuanced effects with things like that.
Your videos are so understandable and amazing, thank you
You're very welcome!
Finally someone that answers all the questions i have before i ask them haha 😍 - Subbed
🤗 Pleasure!
very helpful, hadn't heard of the double colon parameter until now (relatively new to midjourney), thanks!
and also liked this one, also learned something about making more variation with the ::1 !!
thanks a lot :)
You're very welcome! 🙂
This was extremely helpful. Thank you. Totally changed my life with these keyword weights!
That's what I thought ;)
Great tutorial! Thank you so much!
You're very welcome!
Wow, this was super informative, thanks man!
Glad you enjoyed it!
*i've only very recently(ish) started messing about with ai art generator programs and have only experimented with about three of them so far but i really like what the results have been...the randomness has been interesting and personally the distorted/disjointed surreal facial assemblages at times give the images an other worldly aspect that suits the artistic narrative i find appealing...i have found that the order in which i input my prompts also effects the final results as the first prompt is given more weight than the others...*
There seems to be a general consensus that words at the beginning get more wieght than others but I really can't say that I'm convinced this is true.
@@TokenizedAI *easy enough to test this theory...generate an image using a set of prompts then then switch the order around after a few images...maybe even have the same prompt more than once but at a different place such as at the beginning and then near the end*
That was very informative. Thank you.
Very useful guide even without sound ON
That was a fantastic explanation for a newbie like myself, As a teacher, you're one of the best. Looking forward to future vids. Tnx
Thank you very much!
Nicely done tutorial, thanks! How about a tutorial on widely used prompt words for styles. With or without: cyberpunk, psychedelic, necromancer, surreal, vaporwave, etc.
Great suggestion! But just to confirm. You want to know what all these difference styles are and what they look like? What specific "problem" should the video address?
@@TokenizedAI Type of image woks well and doesn't with style prompts.
@@TokenizedAI I think he's talking about excluding and including styles in the prompt to see how better to control the outcomes style specifically.
@@studiovulcan4411 Much clearer now, thanks 🙂
outstanding video. Thank you so so much. You brought clarity to something that was "mooshy" before!
But wait....there's more! 😀 We only scratched the surface of what's possible with weights 😉
Thank you bro. This made things easy
No problem 👍
Amazing and thorough video! Love it. Keep it up
Glad you enjoyed it :)
great video, learned something new. Subscribed.
Glad you enjoyed it 🙂
Super helpful thank you so much 🙏🏼
You're so welcome! And thank you!
So useful, Christian, thank you!
Super helpful, thank you! I've just recently started using Midjourney to create art concepts for my film projects. I've been getting some great results that look very realistic and filmic, but I'm still struggling a lot with more detailed framing attempts - like camera angles, positioning of characters within the frame, foreground and background and so on. I'll keep a lookout on your channel to see if you ever discuss these topics. Thanks once again!
I've found specifying the focal length in mm to be pretty consistent
@@Flamamacue I'll give that a try. Thanks!
There are effectively two ways to achieve what you want to do. You either need a super elaborate prompt and some luck. Or you need to use multiprompts. But those are challenging and not easily explained in a comment section. I will be doing a video on this, rest assured.
@@TokenizedAI excellent, looking forward to your video. Thank you!
Perhaps it could help starting with an image prompt. Take a picture reference roughly in the same angle, framing and composition you want and start generating variations with that as a base.
amazing what difference it makes. thanxs for this awesome trick.🔥🔥🔥
My pleasure 😊
So funny that I recently rated one of the images "hacker fighting a tree" and wondered what the heck that was about 🤣 great explanation for weights 👍
lol, really? 🤣🤣
Thank you so much for this explanation. It is very helpful
Thank you so much! Very helpful in paving a path towards what you are looking for.
Glad it was helpful!
Thanks for this information brother
Pleasure
This video was really helpful, thank you, Christian!
Really nice of you to say that :)
i just try it out and its mindblowing
Great to hear 😁
Just started playing around with this myself. I think commas do still have a place. For example you can add weights to entire sentences and/or single adjectives alike: A majestic black horse on top of a mountain::5, sunset::2, etc etc. In this example it isn't just giving "mountain" a weight of 5 but that entire sentence.
Yeah, but it's not doing that because of the commas. It's doing that because that how weights always work in multiprompts. The weight applies to the full sentence, so everything left of the colons....until it hits another double-colon.
By the way, I use commas too. But primarily to makes things more easy to read. But they don't really do anything.
@@TokenizedAI It’s slightly confusing then, if every word has a default of ::1 (even if it’s not typed) and you give the last keyword a higher value it’s going to effectively make everything to the left of it that higher value. To me it would make sense that a comma could act as a roadblock to cancel out that effect. Because in massive prompts I don’t really wanna go through and assign everything a weight, just a handful of spots as needed to tweak & boost.
I think there a misunderstanding here. The examples I used only had singular keywords, but the "keywords" can be entire phrases as well.
Example:
/imagine a big house::1 spotless blue sky::2 a field of roses::3
The above example is split into 3 segments.
- a big house (weight of 1)
- spotless blue sky (weight of 2)
- a field of roses (weight 3)
Individual words don't have their own weight unless you turn them into a segment by spliting it with ::
Also with the example of the Hacker and Tree, I think the algorithm only applies the weighting to the word preceding it not everything before it.
@@Mimeniia That's incorrect. This is exactly how multiprompts work. Colons separate your prompt into segments and the weights apply to the entire segment, irrespective of whether it's only a single word or an entire phrase.
My man just found prompt weights and is very excited. 😂
Good to hear :P
@@TokenizedAI But honestly this was actually helpful laid out like this as it reminded me just how important and useful they can be and I’ll use it more. I didn’t expect to get as much out of it as when I started the video. I enjoyed it to the extent that I checked I was subscribed! So 👍 ☺️
🤗
excellent material as always congratulations!!
Thanks, appreciate the feedback :)
Thanks for the tips.
Fantastic explanation, thank you so much! Precision is key in all communication with AI.
Thanks so much for the great feedback! :)
SUMMARY of the video by CHATGPT:
The speaker is discussing the importance of prompt structure when using mid-journey text weights. They explain that many people believe that structuring prompts with commas separates them into different concepts, but that this is not true. They provide examples of prompts that are the same despite the use of commas and explain that mid-journey does not pay much attention to punctuation or sentence structure when interpreting prompts. They suggest that a basic rule of thumb is to not rely on commas to separate concepts in a prompt, and instead to use text weights to control the output. The speaker continues by explaining that text weights can be used to control the output of prompts by giving different weights to different keywords or concepts in a prompt. They provide several examples of prompts with weightings and compare them to the same prompts without weightings. They show that the use of weightings results in more creative and varied output. The speaker also explains that text weights can be used with full sentences and that the results can be surprisingly good. They use the example of a hacker fighting with a tree and show that the output is more creative and varied than without the use of weightings.
lol, scary
this is next level scary holy fk
they? Its only one guy
Which service was used to create this concise summary of the transcription/video? Very useful
I am using it and it works well thank you!
Great to hear! :)
Wow! Great tutorial 🔥
Thank you 🙌
Great Prompting tips
Glad it was helpful!
Hi,
Many thanks for sharing this very useful knowledge !
You are welcome! Have fun going wild in Midjourney :)
Midjourney is a godsend for workflows, productivity and creative output, thank you for helping us harness the possibilities even more "precisely" to our imaginings! : )
Just doing my best to deliver what I would want to see :)
Super informative and helpful! Amazing video!!
Glad you think so!
I want to recreate images of Biblical events through Midjourney. But not doing a very good job. Kindly make a video on this.
Thanks for the tips!!!! Great video.
No problem!!
this is fantastic, thanks a lot
You're very welcome!
Great and valuable content, thanks!
You're welcome! 🙂
Really great video, totally unable to thank you . Absolutely amazing.
You're very welcome! The fact that you're commenting is already a great reward for me :)
very interesting ,tks i follow and well explain,we learn !
You are welcome!
this is super helpful, thanks for sharing!
You're so welcome! More to come soon :)
Excellent Tutorial
Glad you think so!
Text weights really give insane results.
I tried it and generated amazing artworks.
Thank you for sharing this.💛🙌
You're welcome 😊 Wait until you see what what else is possible!
Thanks for this!
You're welcome!
Very informative and helpful video. Thank you.
You are welcome!
If you want a hacker fighting with a tree. You should also think about the sentence more. "a hacker fighting a tree" or "a hacker fighting against a tree" would get more consistent results (and less hackers with trees).
Yup, in hindsight, wasn't the best choice of words.
Thanks for these awesome tips! Really helpful.
Glad it was helpful!
Impressive - thanks a lot!
Glad you liked it!
Wow, danke dir :D direkt mal ein abo da gelassen
Das freut mich :) Danke!
Trying hard to do same with your advises.
Dude, thanks so much.
No problem!
River mountain pink.. looking completely same :D
used to prompt "photo, full car composition, photorealistic, motion blur on car rims, motion blur on environment ::2" like a fool ! thanks
Well, I hope the resulting image at least looked TWICE as good as before 😁
Thanks, Super helpful, video::
You're welcome!
great video, thanks!
You're welcome!
Important video. I was doing it wrong. Thank you
Glad it helped 🙂
Good Info. Thanks!
Glad it was helpful!
Great tutorial thank you!
You're very welcome!
Amazing channel!
Thank you!
So useful !
Glad you think so!
Blessings to you!
Thank you!
Mega cool great, thank you ♡
Glad you like it!
Can you add weights to reference image links? I’m trying create a cyberpunk character on horse but struggling to get midjourney to reference my character and horse images equally
In v3 you can use --iw (=image weight) to define the overall weight of the images. However, this parameter isn't supported by v4 yet.
@@TokenizedAI ah hopefully V4 comes out of beta soon then! V3 isn’t to great compared to V4
Nice work!
Thanks!
Subscribed 💯💪🏼
I've found that for pics with lots of inputs and remixes that you have to really crank up the weight to get a palpable change. even up to the max of 10000.
Usually I'll start with small numbers, especially when the prompt is still small. after many remixes it could be very big where I find I'll crank the numbers up as I remix it to keep pushing those aspects of the work that are still lacking.
I've found too that when the prompt is very small, like say just 2 words, that it tends to overpower to one side or the other. if you add some benign additions and give them small weights too it can help to better balance the numbers in those earlier stages of crafting the prompt.
That's because the weight values aren't absolute numbers. They are relative to the overall number of weights you've added. Say you have 4 words with 4 different weights (2,1,4,3). The total weights are 10. So im that context, 3 is actually just 30%.
If you have less words, then it will likely be stronger.
@@TokenizedAI makes sense, ty
Thank you !
You're welcome!
This was so surprisingly educating.
Genuinely. Thank you for making this video!
Thanks for the feedback and glad to hear you enjoyed it. More is coming soon! :)
Very useful. Thank you
You're welcome 🙂
Can you do a tutorial on Blue willow next? I really want to use these tools but right now I cannot afford Mid Journey's prices!
I haven't really used Blue Willow much and if you check Google Trends, the interest in it is nowhere near the same level. I think mainly because it's effectively the same thing. It also uses Stable Diffusion under the hood and copied most of Midjourney's workflow. So I'd assume that they'll also copy most of the parameters as well, which means that same prompts might work in a similar way.
Cool. Thx!
You're welcome!
Nice one my friend thank you.
Does this weighting work with an image url too.
( I guess I should just go and try.. )
Thanks Again for your great videos
There is image weight (--iw) which works in v3. However, that's very different from text weights. And it doesn't work in v4 yet.
@@TokenizedAI Thanks, that's why I couldn't get it to work in v4 then. 😂😂
Listen here you! THANK YOU.
You are so welcome :)
Could perhaps try stuff like
"hacker fighting with::1 fighting with a tree::1"
Great video btw
Thanks!