This is probably the first Midjourney related video over 10 mins that I finished in one sitting. You waste no time and speak concisely, for which we're appreciative.
Thank you! Just subscribed. Your videos are awesome, very detailed and one can tell you are not only passionate, but a great teacher. Looking forward to seeing your channel grow. Thanks for teaching!!
Thanks so much. I honestly never thought of myself as a particularly good teacher. I'm actually quite impatient. But maybe the video format makes it easier for me :D
What about if i keep the same seed as you say but change only the --ar so i can have the same image but in a different ratio aspect (example 9:16) Can i do that ? If not , how can i have the same image in diferent sizes for a post ,for a facebook banner, for instagram story etc?
Doesn't really work. You can try it but it will usually squeeze or stretch the original image to a point where it looks dumb. At least that's how it's behaved for the past few months. Not sure how yesterday's update with the new ratios and slight improvements may have affected this though.
Thank you my brother for the explanation, I benefited from it a lot, the best channel I come across that explains at length. There is a movement that can be used in V4 and it is when you put seed and a specific number after the result is released the same operation can be repeated but by changing the text or adding text to the previous text with the seed mode that you put previously results will appear similar to the previous one, I may have explained this in your clips but I may not have reached it, in general thank you for your effort we are with you.
However, if you use this method the next day it won't work. You need to use the same seed within a very short time to get the same results. They must randomise the seed numbers over time too.
Technically, using the same seed isn't supposed to return the exact same image anyway. It's only meant to produce more consistently "similar" compositions. It's also very possible that Midjourney uses additional inputs to increase the randomness, such as the current date. So, for example, using the same seed may produce very similar results on the same day, but 1 day later it might change. Though this is just a guess. Either way, it's difficult for me to judge without having seen your prompts and images.
@@TokenizedAI I've been using MJ for quite a while now and think they probably randomise a variety of functionality over time as each day, using the same prompt and style and/or artist I get very different results. Which is both good and bad depending on what your end goal is. I'd really like them to have some form of in-and out-painting available like Dall E2. Whatever, for me at least, I find MJ to be head and shoulders above all the other services. They somehow have been able to algorithmise (sic) the entire concept of what we call 'art' like no one else before. Whether or not your prompts produce what you specifically want there's no denying virtually every piece of work it produces has an impossibly high content of 'art'. It's an absolutely enormous achievement and all that work for MJ should be immensely proud of what they've produced...
@@MikeKleinsteuber I agree that some form of in- and outpainting would be great. At the moment, if I need to edit something from a MJ image, I actually do it in DALL-E 2 and then I use an AI upscaler to get a larger version of it. I'm sure they are working on adding this capability to MJ, but I think getting the UI/UX right is a pretty difficult task. Especially from within Discord.
Artificial intelligence (AI) systems are often used to generate images, but these systems generally do not produce the same exact image twice. This is because most AI systems are designed to generate a diverse range of images, rather than producing the same image over and over again. There are several reasons why AI systems might produce different images each time they are used. One reason is that the AI system may be designed to generate randomly generated images, which will naturally be different each time the system is run. Alternatively, the AI system may be trained on a dataset that includes a wide range of different images, and the system is designed to generate images that are similar to the ones it has seen in the training data. In either case, the specific characteristics of the generated images will depend on the specific AI system and the techniques and approaches it is using.
Unfortunately, I'm still trying to figure out how to find a good balance between the amount of time it takes me to create a video and its quality. It's not easy. I'd love to churn out more content faster but I'm afraid of the quality suffering.
Thx for this very useful information on how to use these parameters. The seed paramter can be viewed on one's personal Midjourney homepage, too, where all the jobs are available. And "sameseed" seems to be incompatible with version 4 and 5 of Midjourney.
Great videos, keep up the good work. Tip to save yourself a couple of clicks. Click your mouse's scroll wheel on an image to automatically open it in a new tab. This applies to other places on the web. Click a link with the scroll wheel; and it will open in a new tab for you.
@Tokenized AI by Christian Heidorn like 4 pictures and when you put --sameseed xxxx it always picked first in the 4 pictures box and I am asked how to pick the 2 picture in 4 box
I still struggle to find how to modify only certain part of the seed image, for exemple in his video, with the house, how I can make the sky sunset and keep the exact same house ??
I have followed your instructions, and have tried to use the "seed" in the prompt. I have tried six times but I have received completely different images each time. I have used the copy command to make sure the prompt is the same wording and seed number, but the images a so different. Can you suggest what I am doing wrong.
I can't really help unless I see your exact prompt. It sounds like your either doing something wrong with the parameter or you're hitting the re-roll button, which actually changes the seed every time. Or you're using v3, which has completely different behaviour.
Same here. Wasn't able to reproduce the images. I'm using v4. I found out that there are two seed values. I'm confused. One is the seed value that we specify in the prompt. Once the image is generated and when we request the seed using the envelope, it gives a different seed value. How does that work? Which seed value are we supposed to use in order to regenerate the image? I used both, but every time it was a different image.
@Tokenized AI by Christian Heidorn Thank you, do you have any video showing how to go further editing only one of the images while keeping that image unchanged and while adding more objects to that image? Lets say you generated 4 images and now you like one and you want to work on that one image while keeping it unchanged.
Your videos are very interesting and useful. Advise on the best way to generate pictures with your face in different styles to stay like yourself. maybe even record a video about it, I think many people would be interested.
so --sameseed is like if you could create a chaos value below zero/the default? It would be great if you could tweak it so that the images were even closer variations down to just some differences in pixels.
Thanks so much! I'm definitely still learning the ropes, trying to figure out what works and what doesn't. I know most people recommend super short videos but I kind of prefer to go into detail. As simple as these tools are, mastery is a lot more nuanced.
@@TokenizedAI I,ve personally made over 18,000 images in midjourney in about 5 months and it takes alot of iteration to get good results. I feel i am still at surface level of what midjourney is really capable of.
@@TokenizedAI I,m doing an analysis of the images i've made to come up with best practices from all these experiments, then i'll share with the community.
If I want to create several icons for example and I also want them all to be in the same style, does it help, if make them all the same seed? even though the prompt description changes? If not, how do you use seed? what is the sense of creating exactly the same image once more? What the difference between sameseed an "create variations" in discrod? thanks for your videos! they rock
I don't think Midjourney is currently able to create words or typography that actually makes any sense. The AI model is just trying to create something that "look like" text, but it can't actually create specific sentences or words.
Can you make a video on making the same characters consistantly with a seed number or the best way you know how. Trying to use it to make stories. Also two same characters consistantly in different pics on the same screen. Very good vids so far.
I'm still experimenting with this myself. I don't think that --seed is the best way to do this. My guess is that you'd probably have to use image prompts to achieve this.
Thanks, two part question: If I send my image prompt that has a seed number in it to a client, will he be able retrieve the exact same image with his MJ account? Also I saw somewhere a sameseed parameter, do you think its better to use --sameseed instead of just --seed?
From what I've heard, your client would probably get a different image. It seems that MJ uses additional inputs (possibly the current date, maybe even your user ID) to randomize the results. So while it way work on the same day, it's unlikely it'll work the same way on the next. --sameseed, unlike --seed, will ensure that all 4 variations are very similar to eachother. It kind of depends on what your objective is.
Thanks for the fast reply. I think its defeat the purpose having this seed paramter. I saw a website called promptbase that sells prompts, but what people don't realize is that, they will never be able to reproduce and modify the exact looking image.
I tend to agree. I think the MJ doesn't actually want everything to be 100% reproducible to the tee. --seed is primarily meant to provide the creator with more consistency while working on a prompt. Your particular use case probably isn't within their scope.
I'm not 100% sure what you mean by "very alike", but if you want all 4 variations to look very similar to eachother, then --sameseed is what you need to use.
I see what you mean. You're probably trying to achieve something similar to what Lensa or Avatar AI do. I haven't experimented with that type of use case so far, but I'll look into it. I would assume that you would probably need to use at least 5-6 image prompts in a single command and all images would need to be close ups of the face from different angles. That's how these tools usually work.
Great vid. Even when I use the same seed, MJ adds an additional -s 50 to the image which creates additional changes. How can I turn that additional -s 50 off so I only am using a seed value that I intend? I’ve checked /settings and see that there are only three style options with none being “style off.” Thanks
Do you go over how to apply a seed number to your chosen uploaded image that you just want upscaled, & not changed in any way? Midjourney should have a way for users to be able to simply upscale their uploaded images _without_ any changes to their image! If you have any vids that cover this specific issue, please point me to its link. Thanks for listening.
Great video! How do go on a channel where you’re able create without the mess of other people’s images? It doesn’t look like you have to scroll looking for your images after you create something.
You can, but NOT in v5 because v5 does not have a real upscaler yet. Many people are asking this because they don't understand what's happening in the background.
If you check the links of the channel or the description of the video, there should be a link to a waiting list. I don't have a course right now, but I've been asked this multiple times now, so I will probably have to stop procrastinating and finally create one for beginners and intermediates.
I have a question: How can I tune MidJourney to be able to create a comic book, so a set of pictures, that show very similar characters and settings from picture to picture?
thanks for the video, one thing i'm struggling to understand: whats a reason i would want to use the same seed (in v4) to generate the exact same image?
There is none. However, if you like an image's composition, but want to make some smaller changes to the prompt, then setting the same seed to help you maintain that composition.
@@TokenizedAI I think, being able to reproduce the same image will be very useful. It can function similarly to an address or a url. The whole prompt along with the seed will be like a token for that particular image. This will allow you to own a particular prompt like an NFT. That will increase the value of AI art considerably in my opinion. What do you think?
Does the see work with the 'Variations' or 'Upscaled' image? I tried to send a message and get the seed number but it generated totally different images.
It turns out that the --seed parameter's behaviour is not consistent over time. Apparently it uses something like the current date in its algorithm. So if you use the same seed on the same day, then it might deliver consistent results. But the next day might already be different. Obviously this is a bit annoying and I'm not sure how to fix this.
Thank you for this tutorial. I guess this is what I've been looking for. You see, I want to create a RUclips channel on storytelling and I needed to know if I can generate consistent characters, but in different moods. I think this is just my answer, so thank you. Is there anything I might need though?
Hmm, difficult to say without knowing a bit more about what exactly you'll talk about in your new channel. How exactly to do plan on using Midjourney? Is you objective to use MJ to create illustrations for actual stories that you tell in your videos?
@@TokenizedAI Its more of fantasy type stories. Take superhero comics for example. Something like that, but in my own style obviously. Like images depicting certain actions my character does in that story.
Hey, thanks for this tutorial! Can the seed parameters be used to create a series of different images that share the same style? For example, to illustrate a children’s book?
Good question, don't think so, because --seed primarily controls the composition from what I can tell. You may want to look at how the --chaos parameter works. Low chaos values might help. And you can potentially control the style by providing an image prompt that contains the style you want.
Hello, Is there a way to keep the same color theme across different images? I am working on landscape images but i also want to create a series (for wall art) but I want to keep the same color theme. is this possible? :)
Color schemes are tricky because the more colors you add to a prompt, the more they bleed into different parts of the image. You're probably better off using something like "theme infusion" to achieve a similar result. I'm releasing that video today.
@@TokenizedAI Hello again :) I am really confused on a subject i read about and i am hoping you can help me, I am reading that CMYK is the color mode art should be in because RGB doesn't show what will really be printed? so allot of the AI colors I used for T-shirts are bright, like bright pinks ect... Do you happen to know the color mode of midjourney ? and if I am using prints for instant downloads what color mode should i be in? Thank you!
Another great video, learning a lot with your videos...but I do have a question. Is it possible to have midjourney create characters with different seed numbers, and then later on bring the 2 or more together in the same scene? For example if I'm trying to illustrate a story?
I don't think --seed is the best way to achieve this. I haven't experimented with this yet but I think image prompts are more effective for this sort of thing.
Depends. Upscales have a tendency to add so much detail that your original image may actually end up looking very different in the details and possibly also worse. The recommended solution to this is to try "light upscale redo". But generally, your image should change too drastically.
the use of seeds is very interesting to me. but why have I never seen a seed even though I have followed your tutorial? or is this because it is still in free status? I don't want when the status changes to paid I get the same experience.
I'm not sure I understand your question. Using --seed is a relatively advanced feature. If you're still in your trial and in the #newbie channels, then it's unlikely that you would encounter these.
@@TokenizedAI I would like to change lets say just color of the sky, but keep the exact same house, but each time I try to change ONLY the sky, it change the whole picture and house is totally diferent, so probably it doesnt work in mid journey, best way to do it to use mid journey to find the perfect building, and then adding sky color in photoshop.
@@diniidesign9852 Most likely, you will have to take the original image of the house that you want, upscale it, and then use it as an image prompt to repeat the process. Example - Step 1: "a red house" - Step 2: Upscale your preferred image. - Step 3: Use the image as an image prompt and change your text prompt to "a red house in front of a blue sky::1 a blue sky with a red house::1" But your house won't look exactly the same, no matter what you do.
@@TokenizedAIThank you for the quick reply. please tell me how to solve my problem? when I upload my photo, I add a prompt, I'm not even close to what I want to achieve.
You're using the wrong tool. What you want to do isn't possible with Midjourney. The risk of deep fakes is too big, so Midjourney won't let you do that.
Great video, thanks for sharing, but something doesn't add up to me though. I get a 10 digit seed number for all my images even for the ones I prompted with let's say --seed 1234. Do you guys know why?
Two possible reasons: 1. You re-rolled. When you re-roll a prompt, the underlying seed is randomized again, even if your prompt states otherwise. 2. You retrieved the seed of an upscaled image. When you set a seed in your prompt, it applies to the set of 4 variations. Obviously, 4 separate upscales cannot have the exact same seed.
I have to wait too. I've just cut out the waiting time while editing the video. Regarding the room, check this video: ruclips.net/video/sihp8OSOH6k/видео.html
I really appreciate how good you explain everything, keep up the good work! I have a question though, why would I need the --seed # parameter to retrieve an old work or composition when I can just go back in chat and work on it there? I just don't understand how to make purpose of it.
Good question. Yes, you could go back in your chat an work on it from there. But let's say you want to change something about the prompt and still use the original seed. If you go back in the chat, you can only do that with Remix mode. But Remix mode is known to be quite unreliable. So this gives people the opportunity to simply create a new set of variations with the adjusted prompt. At least in theory.
--sameseed is supported by v4. And "--v3" didn't work because the parameter isn't called "v3". Instead, you need to add the "--v" parameter followed by a "3" as the value --> "--v 3". Note the space.
It depends on what you want to do. If you just want to keep things consistent when entering prompts, then you can set whatever seed you want. The point is to use the same one. If you want to retrieve the seed for a prompt you entered earlier, you can do that by reacting to the image with the envelope icon. You should probably watch this: ruclips.net/video/VEsCj8cMOuo/видео.html
v5 does not have an actual upscaler yet. Therefore there isn't any seed that you could retrieve from an upscaled image. It works as intended for the image grids.
@@TokenizedAI so, apparently it works in v5 if you "envolope" the original 4 imagines generated from the prompt, but it does not work once you selected a single U-version image
Hi Christian, thanks for the information. Very helpful! Got a question. I always get an error when trying to use a seed. It says 'invalid value.' Do you know how to solve this? Has it something to do with the settings? Hope you can help me out.
Doesn't every image you create technically have a seed number even when you don't define one? How would I use a seed from a previous image to generate more like it? I would like to get a seed number from an upscaled image and use that image as a basis for new images. Is that doable?
@@TokenizedAI Kinda sorta not exactly but tbf this is a 3 month old video on an older version of Mid Journey. I wanted to get a seed from a single previously upscaled image to work from. However it seems in V5 you can only retrieve/use seed numbers from the initial 4 image grid image, which was a disappointment. However, after playing around with the -seed parameter I’ve concluded it’s basically useless. And -sameseed is essentially doing the same thing as variations. Or had you selected “V1” in your example. Getting much better results dialing in looks using more robust extensively detailed prompts. When you use short prompts with just a couple words, you going to get more random results because you are leaving a lot of holes that MJ randomly fills.
The video explains it with regards to the basic process. The only reason it doesn't work for your specific case is because MJ hasn't released an actual upscaler for v5 yet. No upscaler, no seed, because there's been no additional processing. Once they do, the process will be the same. This is limitation of v5, not the method. --sameseed also isn't available in v4 and v5. Please refer to the official docs to get clarity on what parameters work in each particular version. docs.midjourney.com/docs/parameter-list
@@TokenizedAI Yeah I noticed V5 images were not really upscaling or I thought they were upscaling the initial 4 instead. But if these are going to be upscaled even higher, that’s insanely great news. Tbh I’ve never used -sameseed just judged it based what I saw it do in your video. It seemed to be a longer way to arrive at what clicking the variation (V1 - V4) buttons would produce. So I would never use that. I’ll try experimenting with -seed and some V4 images. 👍
I operate pretty much exclusively in private mode which is basically communicating direct message with the bot from the get go, rather than through a channel, if I try to find the seed by selecting the envelope on an old image, it does nothing (I guess because I'm already in direct messages)
Sie sind der einzige Experte hier auf RUclips, der auf die Privatsphäreneinstellung hingewiesen hat - Danke !
Freut mich, dass Sie Nutzen aus den Videos ziehen konnten 🙂
This is probably the first Midjourney related video over 10 mins that I finished in one sitting. You waste no time and speak concisely, for which we're appreciative.
And this is actually one of my less concise videos 😬
you make the most useful videos about mj tips and prompts i've found so far. thanks
Awesome, thank you!
I am revisiting this video because it is exactly what I need to know right now. Thanks for all your great work!
Fascinating. The rabbit hole just keeps going deeper and deeper and i love it. Thank you!
And we're only at the very beginning 🙂
thanks for your generosity! be blessed
Amazing thank you! Love your videos
You are so welcome!
Love these videos! Make sure to continue uploading!
Thanks a lot. Considering all the positive feedback I've been getting, I will definitely keep at it :)
My best teacher ,,, thanks for sharing
Thank you! Just subscribed. Your videos are awesome, very detailed and one can tell you are not only passionate, but a great teacher. Looking forward to seeing your channel grow. Thanks for teaching!!
Thanks so much. I honestly never thought of myself as a particularly good teacher. I'm actually quite impatient. But maybe the video format makes it easier for me :D
What about if i keep the same seed as you say but change only the --ar so i can have the same image but in a different ratio aspect (example 9:16) Can i do that ? If not , how can i have the same image in diferent sizes for a post ,for a facebook banner, for instagram story etc?
Doesn't really work. You can try it but it will usually squeeze or stretch the original image to a point where it looks dumb. At least that's how it's behaved for the past few months.
Not sure how yesterday's update with the new ratios and slight improvements may have affected this though.
Thank you my brother for the explanation, I benefited from it a lot, the best channel I come across that explains at length.
There is a movement that can be used in V4 and it is when you put seed and a specific number after the result is released the same operation can be repeated but by changing the text or adding text to the previous text with the seed mode that you put previously results will appear similar to the previous one, I may have explained this in your clips but I may not have reached it, in general thank you for your effort we are with you.
Yes, I am aware 😉
boooommm subscribed, quick to the point less fancy pictures that we can already see inside MJ, more straight up direct useful tips
Appreciate the kind feedback! :)
I subbed because you're clear and easy to understand, comprehensible, and you time code your videos.
Appreciate the feedback. It's great to hear that many people are interested in more thorough and detailed explanations 🙂
@@TokenizedAI Cool, many channels are not so clear or easy to understand.
However, if you use this method the next day it won't work. You need to use the same seed within a very short time to get the same results. They must randomise the seed numbers over time too.
Technically, using the same seed isn't supposed to return the exact same image anyway. It's only meant to produce more consistently "similar" compositions. It's also very possible that Midjourney uses additional inputs to increase the randomness, such as the current date. So, for example, using the same seed may produce very similar results on the same day, but 1 day later it might change. Though this is just a guess. Either way, it's difficult for me to judge without having seen your prompts and images.
@@TokenizedAI I've been using MJ for quite a while now and think they probably randomise a variety of functionality over time as each day, using the same prompt and style and/or artist I get very different results. Which is both good and bad depending on what your end goal is. I'd really like them to have some form of in-and out-painting available like Dall E2. Whatever, for me at least, I find MJ to be head and shoulders above all the other services. They somehow have been able to algorithmise (sic) the entire concept of what we call 'art' like no one else before. Whether or not your prompts produce what you specifically want there's no denying virtually every piece of work it produces has an impossibly high content of 'art'. It's an absolutely enormous achievement and all that work for MJ should be immensely proud of what they've produced...
@@MikeKleinsteuber I agree that some form of in- and outpainting would be great. At the moment, if I need to edit something from a MJ image, I actually do it in DALL-E 2 and then I use an AI upscaler to get a larger version of it. I'm sure they are working on adding this capability to MJ, but I think getting the UI/UX right is a pretty difficult task. Especially from within Discord.
@@TokenizedAI I've tried that method but for the sort of stuff I do Dall E2 can't hack it and I really don't like the images it produces
Artificial intelligence (AI) systems are often used to generate images, but these systems generally do not produce the same exact image twice. This is because most AI systems are designed to generate a diverse range of images, rather than producing the same image over and over again. There are several reasons why AI systems might produce different images each time they are used. One reason is that the AI system may be designed to generate randomly generated images, which will naturally be different each time the system is run. Alternatively, the AI system may be trained on a dataset that includes a wide range of different images, and the system is designed to generate images that are similar to the ones it has seen in the training data. In either case, the specific characteristics of the generated images will depend on the specific AI system and the techniques and approaches it is using.
First video I've seen of yours. Great value. Thank you
Your videos are so full of helpful information. You are an excellent instructor and I really enjoy your teaching style.
Thank you so much! I'm still learning the ropes myself 🙂
Love your videos! Please continue uploading... I need more of them! Best regards from Poland.
Unfortunately, I'm still trying to figure out how to find a good balance between the amount of time it takes me to create a video and its quality. It's not easy. I'd love to churn out more content faster but I'm afraid of the quality suffering.
Thx for this very useful information on how to use these parameters. The seed paramter can be viewed on one's personal Midjourney homepage, too, where all the jobs are available. And "sameseed" seems to be incompatible with version 4 and 5 of Midjourney.
Because there is no version 6? 😉
@@TokenizedAI The Curse of the Typo 😆
Your videos have been amazingly helpful. Thank You! Waiting for the next one to drop!
Working on it as we speak!
Great videos, keep up the good work. Tip to save yourself a couple of clicks. Click your mouse's scroll wheel on an image to automatically open it in a new tab. This applies to other places on the web. Click a link with the scroll wheel; and it will open in a new tab for you.
Except that my Apple Magic Mouse 2 doesn't have a "scroll wheel", so I'm not really sure that mechanism works for me. 😁
13:34 how I can pick the second picture for same seed?
What exactly do you mean by "second picture"?
@Tokenized AI by Christian Heidorn like 4 pictures and when you put --sameseed xxxx it always picked first in the 4 pictures box and I am asked how to pick the 2 picture in 4 box
Upscale the image and then get the job details of the upscaled image.
@@TokenizedAI seed number or job ID?
You need to the seed. Your image is a job. The "job details" include job ID, seed and basic prompt.
Hello from France, I love the content of your website and your youtube videos.
Merci beaucoup! Really appreciate all the nice feedback :) Messages like this motivate me to keep going!
So much value and clarity. Thank you!
Solid video! i found the part regarding using the same seed valuable to be extremely helpful.
Glad it was helpful! Some like to use --seed, others don't. It has it's benefits in certain situations.
I still struggle to find how to modify only certain part of the seed image, for exemple in his video, with the house, how I can make the sky sunset and keep the exact same house ??
I have followed your instructions, and have tried to use the "seed" in the prompt. I have tried six times but I have received completely different images each time. I have used the copy command to make sure the prompt is the same wording and seed number, but the images a so different. Can you suggest what I am doing wrong.
I can't really help unless I see your exact prompt. It sounds like your either doing something wrong with the parameter or you're hitting the re-roll button, which actually changes the seed every time. Or you're using v3, which has completely different behaviour.
Same here. Wasn't able to reproduce the images. I'm using v4.
I found out that there are two seed values. I'm confused.
One is the seed value that we specify in the prompt.
Once the image is generated and when we request the seed using the envelope, it gives a different seed value. How does that work? Which seed value are we supposed to use in order to regenerate the image? I used both, but every time it was a different image.
Finally found a decent explanation, thanks😀
You're welcome 🙂
thank you so much for such an indepth video, would you do another one on how to make the same/similar characters do different actions?
Please check my channel. I've done those already
Thanks for this tutorial. I'm not able to activate the direct message option setting (15:11). I have the paid acct. $30/month. Any recommendations?
It doesn't what plan you have. It should even work with a trial. What exactly is preventing you at that specific spot. Is it disabled?
Boom once again you changer my midjourney World !!!
Thanks for the kind words :)
great new channel!! thanks for your effort! 🙏👍🏻
Appreciate the nice feedback 🙂
On-point tutorial, thanks a million.
My pleasure
Thank you for detailed and easy explanation! Definately you have a talent for teaching!
Glad it was helpful! I think some of my videos may be too detailed for advanced users but I think it helps novices quite a bit.
Excellent Channel. I've subbed
Thanks for the sub!
@Tokenized AI by Christian Heidorn Thank you, do you have any video showing how to go further editing only one of the images while keeping that image unchanged and while adding more objects to that image? Lets say you generated 4 images and now you like one and you want to work on that one image while keeping it unchanged.
Unfortunately, I don't.
Wow. This is amazing. Thank you!
thank you again from Berlin, germany. #this helps a lot
My pleasure
Your videos are very interesting and useful. Advise on the best way to generate pictures with your face in different styles to stay like yourself. maybe even record a video about it, I think many people would be interested.
Just add 4-5 images of yourself (your face) from various different angles as image prompts. That usually does the trick.
@@TokenizedAI can you write something for example?
so --sameseed is like if you could create a chaos value below zero/the default?
It would be great if you could tweak it so that the images were even closer variations down to just some differences in pixels.
Not exactly, but it's similar in outcome.
Great video, very well explained. Wish you best of luck in your RUclips Journey. Subbed!
Thanks so much! I'm definitely still learning the ropes, trying to figure out what works and what doesn't. I know most people recommend super short videos but I kind of prefer to go into detail. As simple as these tools are, mastery is a lot more nuanced.
@@TokenizedAI I,ve personally made over 18,000 images in midjourney in about 5 months and it takes alot of iteration to get good results. I feel i am still at surface level of what midjourney is really capable of.
@@pichacraft Yikes! 18,000 images a quite the tally, haha. But yeah, totally agree that we're mostly just scratching the surface.
@@TokenizedAI I,m doing an analysis of the images i've made to come up with best practices from all these experiments, then i'll share with the community.
Thank you!! You're a great teacher!
Thank you so much :)
really well explained, thank you :)
Glad it was helpful!
Thanks a lot. How do you organize your Discord surface, so that it only shows your own work???
I think you'll want to watch this video: ruclips.net/video/sihp8OSOH6k/видео.html
Amazing job! Specially the seed trick at the end :)
Btw, you have a typo around 11:22 the prompt showed on screen does not match the one you used.
Good catch! I must have forgotten to update the full prompt. Sorry about that 😬
Awesome dude!!
Thanks for watching!
If I want to create several icons for example and I also want them all to be in the same style, does it help, if make them all the same seed? even though the prompt description changes? If not, how do you use seed? what is the sense of creating exactly the same image once more? What the difference between sameseed an "create variations" in discrod? thanks for your videos! they rock
The seed doesn't determine the style. It mostly just impacts composition.
Please do a video on the most difficult part of Midjourney "Making Words". Thank you
I don't think Midjourney is currently able to create words or typography that actually makes any sense. The AI model is just trying to create something that "look like" text, but it can't actually create specific sentences or words.
Thanks bro. Good tutorial.
Glad it helped :)
Servus und danke fuer das Tutorial
Can you make a video on making the same characters consistantly with a seed number or the best way you know how. Trying to use it to make stories. Also two same characters consistantly in different pics on the same screen. Very good vids so far.
I'm still experimenting with this myself. I don't think that --seed is the best way to do this. My guess is that you'd probably have to use image prompts to achieve this.
Thanks, two part question: If I send my image prompt that has a seed number in it to a client, will he be able retrieve the exact same image with his MJ account? Also I saw somewhere a sameseed parameter, do you think its better to use --sameseed instead of just --seed?
From what I've heard, your client would probably get a different image. It seems that MJ uses additional inputs (possibly the current date, maybe even your user ID) to randomize the results. So while it way work on the same day, it's unlikely it'll work the same way on the next. --sameseed, unlike --seed, will ensure that all 4 variations are very similar to eachother. It kind of depends on what your objective is.
Thanks for the fast reply. I think its defeat the purpose having this seed paramter. I saw a website called promptbase that sells prompts, but what people don't realize is that, they will never be able to reproduce and modify the exact looking image.
I tend to agree. I think the MJ doesn't actually want everything to be 100% reproducible to the tee. --seed is primarily meant to provide the creator with more consistency while working on a prompt. Your particular use case probably isn't within their scope.
wonderful elaboration
Thank you
Hello, what prompt should i use to make the faces look very alike when going from an image to a realistic ai image?
I'm not 100% sure what you mean by "very alike", but if you want all 4 variations to look very similar to eachother, then --sameseed is what you need to use.
@@TokenizedAI i mean same face features as the picture i feed it
I see what you mean. You're probably trying to achieve something similar to what Lensa or Avatar AI do. I haven't experimented with that type of use case so far, but I'll look into it. I would assume that you would probably need to use at least 5-6 image prompts in a single command and all images would need to be close ups of the face from different angles. That's how these tools usually work.
@@TokenizedAI exactly, yess i have tried some had best succes with ultra realistic in --v 4 but still it s making the facial treats too sharp
You can try using a lower quality level. Or, if the problem is with your upscales, rather use Light Upscale than the regular one.
awesome video!
Thanks!
Great vid. Even when I use the same seed, MJ adds an additional -s 50 to the image which creates additional changes. How can I turn that additional -s 50 off so I only am using a seed value that I intend? I’ve checked /settings and see that there are only three style options with none being “style off.” Thanks
Change your setting to the medium level. That should apply a stylize value of 100 (if you're using v4), which is the default.
Well explained. thanks a lot!
Glad it was helpful!
Do you go over how to apply a seed number to your chosen uploaded image that you just want upscaled, & not changed in any way?
Midjourney should have a way for users to be able to simply upscale their uploaded images _without_ any changes to their image!
If you have any vids that cover this specific issue, please point me to its link. Thanks for listening.
What you want to do simply isn't possible in Midjourney. That's not how seeds work in this context.
Great video! How do go on a channel where you’re able create without the mess of other people’s images? It doesn’t look like you have to scroll looking for your images after you create something.
Watch this: ruclips.net/video/sihp8OSOH6k/видео.html
Can you get a seed for a specific upscaled image? I tried but can not. How do you direct MJ to work off a specific 1 photo out of 4?
You can, but NOT in v5 because v5 does not have a real upscaler yet. Many people are asking this because they don't understand what's happening in the background.
Hello! I clicked on envelope to get old's job seeds value, but didn't receive the message. Yes the Direct Message was activated from beginning.
You're trying to do this in v5 on an individual upscaled image, which isn't possible because v5 doesn't have an upscaler yet.
Great work! Do you offer courses? I found your explanations very easy to understand.
there is no work, no skills. no one of us who has any will not use this application.
If you check the links of the channel or the description of the video, there should be a link to a waiting list. I don't have a course right now, but I've been asked this multiple times now, so I will probably have to stop procrastinating and finally create one for beginners and intermediates.
I have a question: How can I tune MidJourney to be able to create a comic book, so a set of pictures, that show very similar characters and settings from picture to picture?
Can really explain this in a comment. But I'm working on something. I'll share it soon.
Where can I find a seed chart. How do I know which number to use?
you have a very pleasant voice :)
Thank you 🙂 Pretty sure that's thanks to the Adobe audio enhancer though 😅
thanks for the video, one thing i'm struggling to understand: whats a reason i would want to use the same seed (in v4) to generate the exact same image?
There is none. However, if you like an image's composition, but want to make some smaller changes to the prompt, then setting the same seed to help you maintain that composition.
@@TokenizedAI I think, being able to reproduce the same image will be very useful. It can function similarly to an address or a url. The whole prompt along with the seed will be like a token for that particular image. This will allow you to own a particular prompt like an NFT. That will increase the value of AI art considerably in my opinion. What do you think?
Does the see work with the 'Variations' or 'Upscaled' image? I tried to send a message and get the seed number but it generated totally different images.
It turns out that the --seed parameter's behaviour is not consistent over time. Apparently it uses something like the current date in its algorithm. So if you use the same seed on the same day, then it might deliver consistent results. But the next day might already be different. Obviously this is a bit annoying and I'm not sure how to fix this.
Thank you for this tutorial. I guess this is what I've been looking for. You see, I want to create a RUclips channel on storytelling and I needed to know if I can generate consistent characters, but in different moods. I think this is just my answer, so thank you.
Is there anything I might need though?
Hmm, difficult to say without knowing a bit more about what exactly you'll talk about in your new channel. How exactly to do plan on using Midjourney? Is you objective to use MJ to create illustrations for actual stories that you tell in your videos?
@@TokenizedAI Its more of fantasy type stories. Take superhero comics for example. Something like that, but in my own style obviously. Like images depicting certain actions my character does in that story.
@@peacedessy109 In that case, you may want to look into the --chaos parameter as well, as it can help you reduce randomness even further.
Hey, thanks for this tutorial! Can the seed parameters be used to create a series of different images that share the same style? For example, to illustrate a children’s book?
Good question, don't think so, because --seed primarily controls the composition from what I can tell. You may want to look at how the --chaos parameter works. Low chaos values might help. And you can potentially control the style by providing an image prompt that contains the style you want.
Hello, Is there a way to keep the same color theme across different images? I am working on landscape images but i also want to create a series (for wall art) but I want to keep the same color theme. is this possible? :)
Color schemes are tricky because the more colors you add to a prompt, the more they bleed into different parts of the image. You're probably better off using something like "theme infusion" to achieve a similar result. I'm releasing that video today.
@@TokenizedAI Hello again :) I am really confused on a subject i read about and i am hoping you can help me, I am reading that CMYK is the color mode art should be in because RGB doesn't show what will really be printed? so allot of the AI colors I used for T-shirts are bright, like bright pinks ect... Do you happen to know the color mode of midjourney ? and if I am using prints for instant downloads what color mode should i be in? Thank you!
8888 is a number associated with wealth, money, and good fortune. Thanks for the good luck great resources!
Hey great videos , how do you get your own feed/channel on discord so not having to use newbie feed. I’m presuming it’s part of the subscription??
You need to add the Discord bot to your own Discord server. Check out this video: ruclips.net/video/sihp8OSOH6k/видео.html
Thank you!
Thank you very much bro!!!
No worries. Glad you enjoyed it.
Another great video, learning a lot with your videos...but I do have a question. Is it possible to have midjourney create characters with different seed numbers, and then later on bring the 2 or more together in the same scene? For example if I'm trying to illustrate a story?
I don't think --seed is the best way to achieve this. I haven't experimented with this yet but I think image prompts are more effective for this sort of thing.
How can I get back the old styles. It seems mj updated last weds. All my prompts Ives been entering now are getting lame results now
Just don't use v5 and your problem is solved.
@@TokenizedAI BAD BAD BAD SOLUTION
ok so if u use seed can you say make a building and hav the same building wiz different angles?
Not necessarily the same building, but MJ will use a similar composition.
excellent, thank you!
You're welcome!
Great video But how do you choose the first seed to start?
It's irrelevant. Take whatever you want. The point is just to keep it constant.
@@TokenizedAI ah ok thank' s... From italy
I noticed that if I upscale an image from a few day before it gets pretty distorted, is it bc seeds expire?
Depends. Upscales have a tendency to add so much detail that your original image may actually end up looking very different in the details and possibly also worse. The recommended solution to this is to try "light upscale redo". But generally, your image should change too drastically.
the use of seeds is very interesting to me. but why have I never seen a seed even though I have followed your tutorial? or is this because it is still in free status? I don't want when the status changes to paid I get the same experience.
I'm not sure I understand your question. Using --seed is a relatively advanced feature. If you're still in your trial and in the #newbie channels, then it's unlikely that you would encounter these.
Great video. How I can change only the sky without to change the house in my seed image? thx a lot
Depends on how you want to change it. Do you want to change the color or add clouds or something else? Just add it to your prompt and try it out.
@@TokenizedAI I would like to change lets say just color of the sky, but keep the exact same house, but each time I try to change ONLY the sky, it change the whole picture and house is totally diferent, so probably it doesnt work in mid journey, best way to do it to use mid journey to find the perfect building, and then adding sky color in photoshop.
@@diniidesign9852 Most likely, you will have to take the original image of the house that you want, upscale it, and then use it as an image prompt to repeat the process.
Example
- Step 1: "a red house"
- Step 2: Upscale your preferred image.
- Step 3: Use the image as an image prompt and change your text prompt to "a red house in front of a blue sky::1 a blue sky with a red house::1"
But your house won't look exactly the same, no matter what you do.
@@TokenizedAI thanks a lot for advice! Will try!
hello everyone, does anyone to know how i can get seeds from my uploaded photo. I need exact my own design? Thank you
That's not possible and that's also not what seeds are for.
@@TokenizedAIThank you for the quick reply. please tell me how to solve my problem? when I upload my photo, I add a prompt, I'm not even close to what I want to achieve.
You're using the wrong tool. What you want to do isn't possible with Midjourney. The risk of deep fakes is too big, so Midjourney won't let you do that.
Great video
Thanks!
Does sameseed and --chaos 0 works identical?
Not really.
Great video, thanks for sharing, but something doesn't add up to me though. I get a 10 digit seed number for all my images even for the ones I prompted with let's say --seed 1234.
Do you guys know why?
Two possible reasons:
1. You re-rolled. When you re-roll a prompt, the underlying seed is randomized again, even if your prompt states otherwise.
2. You retrieved the seed of an upscaled image. When you set a seed in your prompt, it applies to the set of 4 variations. Obviously, 4 separate upscales cannot have the exact same seed.
@@TokenizedAI Thanks for the response, it must have been that, and it makes sense in a way.
try to do different types of angles from the same scene. Thanks.
Will take it into consideration.
Can you extrapolate the seed value on an image where you didn't originally specify it?
NM just watch the whole video dum-dum
I see you figured it out yourself. 😁
How do you have your own room and dont wait at all? I need to learn that please.
I have to wait too. I've just cut out the waiting time while editing the video.
Regarding the room, check this video: ruclips.net/video/sihp8OSOH6k/видео.html
@@TokenizedAI thanks for your reply, amazing videos. have a nice day!
I really appreciate how good you explain everything, keep up the good work!
I have a question though, why would I need the --seed # parameter to retrieve an old work or composition when I can just go back in chat and work on it there? I just don't understand how to make purpose of it.
Good question. Yes, you could go back in your chat an work on it from there. But let's say you want to change something about the prompt and still use the original seed. If you go back in the chat, you can only do that with Remix mode. But Remix mode is known to be quite unreliable. So this gives people the opportunity to simply create a new set of variations with the adjusted prompt. At least in theory.
Could not get --sameseed to work, kept saying ""--version 4 is not compatible with --sameseed", did --v3 still did not work Help!
--sameseed is supported by v4. And "--v3" didn't work because the parameter isn't called "v3". Instead, you need to add the "--v" parameter followed by a "3" as the value --> "--v 3". Note the space.
I created a character that renders almost the same everytime, but I cannot give It any objects to hold like a dumb bell for example.
Any ideas why?
Perhaps watch my character design series
@@TokenizedAI thanks
God blessed me to find this channel with infinite value. Thank you so much for your job, sir
Thanks!
I not understand. Where i found this seed ? I need create after use this same code ? Thank you.
It depends on what you want to do. If you just want to keep things consistent when entering prompts, then you can set whatever seed you want. The point is to use the same one.
If you want to retrieve the seed for a prompt you entered earlier, you can do that by reacting to the image with the envelope icon. You should probably watch this: ruclips.net/video/VEsCj8cMOuo/видео.html
this seems to not work for anything created by v5. It works for v4 and anything else, but I am not able to get this to work in any v5 image.
v5 does not have an actual upscaler yet. Therefore there isn't any seed that you could retrieve from an upscaled image. It works as intended for the image grids.
@@TokenizedAI so, apparently it works in v5 if you "envolope" the original 4 imagines generated from the prompt, but it does not work once you selected a single U-version image
Hi Christian, thanks for the information. Very helpful! Got a question. I always get an error when trying to use a seed. It says 'invalid value.' Do you know how to solve this? Has it something to do with the settings? Hope you can help me out.
Sounds like you're entering something incorrectly (like a space in between or something)
does sameseed still not work in v4 until now?
Nope
love your videos (just getting really distracted by the music..)
My newer videos have much less music.
Do not do it. No music at all.
Doesn't every image you create technically have a seed number even when you don't define one? How would I use a seed from a previous image to generate more like it? I would like to get a seed number from an upscaled image and use that image as a basis for new images. Is that doable?
Umm, I kinda explain all that precisely in this video.
@@TokenizedAI Kinda sorta not exactly but tbf this is a 3 month old video on an older version of Mid Journey. I wanted to get a seed from a single previously upscaled image to work from. However it seems in V5 you can only retrieve/use seed numbers from the initial 4 image grid image, which was a disappointment. However, after playing around with the -seed parameter I’ve concluded it’s basically useless. And -sameseed is essentially doing the same thing as variations. Or had you selected “V1” in your example. Getting much better results dialing in looks using more robust extensively detailed prompts. When you use short prompts with just a couple words, you going to get more random results because you are leaving a lot of holes that MJ randomly fills.
The video explains it with regards to the basic process.
The only reason it doesn't work for your specific case is because MJ hasn't released an actual upscaler for v5 yet. No upscaler, no seed, because there's been no additional processing. Once they do, the process will be the same. This is limitation of v5, not the method.
--sameseed also isn't available in v4 and v5. Please refer to the official docs to get clarity on what parameters work in each particular version.
docs.midjourney.com/docs/parameter-list
@@TokenizedAI Yeah I noticed V5 images were not really upscaling or I thought they were upscaling the initial 4 instead. But if these are going to be upscaled even higher, that’s insanely great news.
Tbh I’ve never used -sameseed just judged it based what I saw it do in your video. It seemed to be a longer way to arrive at what clicking the variation (V1 - V4) buttons would produce. So I would never use that. I’ll try experimenting with -seed and some V4 images. 👍
How do you use seed when you are already working privately with the direct message function?
I'm not sure I understand the question. It works the exact same way.
I operate pretty much exclusively in private mode which is basically communicating direct message with the bot from the get go, rather than through a channel, if I try to find the seed by selecting the envelope on an old image, it does nothing (I guess because I'm already in direct messages)
Does anyone know how I can make my own server like this so I dont have to use midjourney's character room? Please advise thanks
Watch this: ruclips.net/video/sihp8OSOH6k/видео.html