Midjourney's Amazing New Command - Diving into Describe / Prompts / Tokens!

Рет қаралды 59,888

Күн бұрын

Today we're taking a look at an amazing new Midjourney Command: Describe-- where Midjourney will actually tell YOU what IT sees!
And even as amazing as that is, there is a lot of hidden details about how Midjourney views images and prompts, so we'll be diving into Midjourney's output prompts to examine what we can discover. Plus, we'll be taking a look at how Tokens work in Midjourney, with some tips on how to maximize your prompt and get the most out of them!
FREE VIDEO HANDOUT: theoreticallymedia.gumroad.co...
AFFILATE LINKS:
Camera: amzn.to/3yXMDY2
Microphone: amzn.to/3K1jIZm
Audio Interface: amzn.to/3lDX9kf
Coffee: amzn.to/3JZuBeq
Follow Me on Twitter: / theomediaai
-------------------------------------------------
Thanks for watching Theoretically Media! I cover a large range of topics here on the Creative AI space, Technology, Tutorials, and Reviews! Please enjoy your time here, and subscribe!
Your comments mean a LOT to me, and I read and try to respond to every one of them, so please do drop any thoughts, suggestions, questions, or topic requests!

Пікірлер: 158

@TheoreticallyMedia Жыл бұрын

You can check out the Free Video Cheatsheet/Handout here: theoreticallymusic.gumroad.com/l/describe

@LouisGedo Жыл бұрын

👋

@stephenrider6107 9 ай бұрын

It looks like you're not on gumroad any more?

@Justin_Hikes Жыл бұрын

You are creating some of the most useful Midjourney vids out there. Keep it up! Thanks!

@TheoreticallyMedia Жыл бұрын

Oh man, that means a lot! Thank you so much! Legit just made my day!

@BitBoy_Gaming Жыл бұрын

Love the vid! Definitely liked the explanation for the tokens! Looking forward to more!

@TheoreticallyMedia Жыл бұрын

Thank you! Yeah, I tend to think most stay away from talking about tokens, since it is a bit mysterious in terms of how MJ uses them. And while I admit that I don’t have the full answer, I think it’s still worth learning!

@TheRino54 Жыл бұрын

Wow, the describe new prompt for MidJourney looks amazing and inspiring! I love how it allows for even more creativity and flexibility in generating text. Thanks for sharing this development with us!

@TheoreticallyMedia Жыл бұрын

Happy to hear! And stay tuned, I have some interesting experiments with describe coming up this week!

@gfranks007 Жыл бұрын

Tim thanks again for telling us about this new feature and breaking it down for us! Insightful and helpful!

@TheoreticallyMedia Жыл бұрын

Thank you, Graeme!! Super happy to hear you're enjoying these videos!

@theblowupdollsmusic Жыл бұрын

Thank you so much for always taking the time to share all these sophisticated mid journey techniques with us. Your content is so thorough and inspiring. It's greatly appreciated.

@TheoreticallyMedia Жыл бұрын

Oh, so happy to hear! Thank YOU!!

@RalphMoreau 4 ай бұрын

Always appreciate your research Tim thank you. As a word nerd this does have some interesting implications.

@teolita Жыл бұрын

I love this new command for MJ, I tested it already and the results were great. Thank you for the in depth video 💯

@TheoreticallyMedia Жыл бұрын

Thank you so much for the comment! You might want to check out the latest video on Photobashing in MJ, I do some more with /describe there as well! And I totally agree, I LOVE this new command!

@evellymartinez31 Жыл бұрын

GRACIAS ! GRACIAS GRACIAS ! you don't have idea how i love your videos , thank you for sharing your knowledge wish you a lot of success in your live 💖

@TheoreticallyMedia Жыл бұрын

De Nada!! Happy you are enjoying and learning! ¡Salud!

@dirksabbe-vf2rv Жыл бұрын

Thanks Tim! Great video as usual! I really like your thoughtful approach to this whole AI thing.

@TheoreticallyMedia Жыл бұрын

Thanks so much! And really, your comment on the approach is highly appreciated!

@jonathoncarter Жыл бұрын

Much appreciated. Your videos are superb!

@TheoreticallyMedia Жыл бұрын

Thank you so much! Very appreciated!!

@savagerey5281 Жыл бұрын

love love love this video and how it makes the picture 4 more pictures :)

@sherry450 Жыл бұрын

Thanks, Tim. Lots of great tips!

@TheoreticallyMedia Жыл бұрын

Thank you so much!! Super happy it was helpful!!

@mhgust Жыл бұрын

Thanks for the great tips Tim!

@TheoreticallyMedia Жыл бұрын

my pleasure!!

@ocelotcat Жыл бұрын

Great video, so interesting. Thank you 🙏

@TheoreticallyMedia Жыл бұрын

Thank you for watching and the nice comment!!

@DAOalfonso Жыл бұрын

Amazing!. Thank you Sir.

@JudiJamiesonThePsychic Жыл бұрын

This was very helpful. Thank you

@TheoreticallyMedia Жыл бұрын

Fantastic to hear! Truly, the comment is super appreciated!

@Digitalnomad8 Жыл бұрын

Great video wow a lot of information. Keep sharing subscribed and liked. Thanks again for posting

@TheoreticallyMedia Жыл бұрын

Thank you so much for checking it out and the sub! Looking forward to seeing you again! (I mean, sorta. You know what I mean! ha!)

@danielsomerstein9996 Жыл бұрын

Best video I have found for Midjourney, back engineering baby!

@TheoreticallyMedia Жыл бұрын

Thank you! Don't know if you caught it, but I just posted a video on a pretty interesting way to combine /describe with ChatGPT. I'm not a huge fan of that whole "Have ChatGPT write a prompts for me" trend, but I think this is a super interesting and creative technique: kzfaq.info/get/bejne/bdqkddl-nc-dj6M.html

@ekarongsukruthaiworakul7720 Жыл бұрын

Very cool and beneficial video. Thanks for sharing

@TheoreticallyMedia Жыл бұрын

Thank you for watching!! Glad it was useful for you!

@Barncore Жыл бұрын

Fascinating. Thanks for sharing

@TheoreticallyMedia Жыл бұрын

Thank you! and thank you for watching!

@classic_dude Жыл бұрын

Thanks, Tim. Well done.

@TheoreticallyMedia Жыл бұрын

Thank you!! And thank you for the comment! Really means a lot!

@thecatslastword7782 Жыл бұрын

This was ridiculously helpful!

@TheoreticallyMedia Жыл бұрын

That's fantastic to hear! Next week I've got a follow up that I think you''ll be interested in as well!

@secretarchivesofthevatican Жыл бұрын

Super useful - thank you!

@TheoreticallyMedia Жыл бұрын

Awesome to hear!! Thank you for the comment!!

@mrschneebly85 Жыл бұрын

Learned a lot thanks!!!❤

@TheoreticallyMedia Жыл бұрын

Your Welcome! And I'm super happy to hear that!!

@TheoreticallyMedia Жыл бұрын

You're Welcome....ugh. Need more coffee!! Haha!

@jodirisenmay Жыл бұрын

Thank you! Newbie here! I've been having fun with moments of frustration when I can't figure out a prompt but I'm devouring all this great info! The fabric is pronounced buh-teek, which is as unique as ai creations! 😊

@TheoreticallyMedia Жыл бұрын

That's great to hear! If you ever get stumped, feel free to reach out! And thank you for the correct pronunciation! I actually ran it a few times in MJ just to see the results, it is really cool! I've seen the style a lot, particularly in my more bohemian musician years, but I never knew what it was actually called! And now, thanks to you, I can actually say it without sounding like an idiot! ha!

@jodirisenmay Жыл бұрын

@@TheoreticallyMedia 😄

@FidelHimself Жыл бұрын

Very interesting!

@TheoreticallyMedia Жыл бұрын

The more I think about it, the cooler it is!

@blacksketchtv7868 Жыл бұрын

thx for u re work

@TheoreticallyMedia Жыл бұрын

You’re welcome! Thank you for the watch and the comment! Much appreciated!

@chrisklotz2137 Жыл бұрын

Thanks!

@TheoreticallyMedia Жыл бұрын

Hi, Chris! Thank you so much for the Super Thanks! That's amazing!!

@LepoppetArtCrafts Жыл бұрын

Interest to see the comparison between "pasting the image link" in the prompt vs "seed XXXXXX" vs "describe". Lots of adventures with this new command.

@TheoreticallyMedia Жыл бұрын

It’s great to see Midjourney starting to push the envelope a bit! I was getting a little concerned they were coasting a few months ago, but I’m happy to see that radio silence was actually grinding!

@MicheleSherwood-cc4xn 11 ай бұрын

Batik is pronounced bah-teek with emphasis on the second syllable FYI . Love your videos!

@TheoreticallyMedia 11 ай бұрын

Thank you!! And Thank you!!

@Queenwraith Жыл бұрын

Great vid as always! Can you use describe prompt to use a source image of eg: a model in front of a plain background, then get it to add in backgrounds you describe?

@TheoreticallyMedia Жыл бұрын

So, I think your best bet there is to Photobash you image. If you check out the channel, the latest video (or second to latest, depending on when you see this) has a tutorial on photobashing that I think will help!

@kiaidrawingss Жыл бұрын

Those images were all so great. Is there anywhere where we can see or download those images you generated?

@TheoreticallyMedia Жыл бұрын

If people want, I can upload the images to a google drive. What I really need to do is finally get around to just uploading all the tutorial images to Instagram. Ha, I think you might have finally inspired me to do that! If there’s a specific image you’re looking for, let me know and I’ll post it up for you!

@mikehuntishuge Жыл бұрын

Thanks.

@TheoreticallyMedia Жыл бұрын

No prob!

@ricksegalCanada Жыл бұрын

These videos are outstanding and super helpful. Death Wish Coffee? You speak really slowly for a guy drinking that stuff. 😉😉

@TheoreticallyMedia Жыл бұрын

Ha! I love it!! But, I do limit myself to only one cup of the Death Wish a day! I mean, I have more Coffee for sure-- I just mellow back to the regular stuff!

@duesenberger Жыл бұрын

Thx for this video. It is really helpful. A question: How can I get consistent images of characters out of Midjourney, if f.e. I want to use it in a novel with the same characters in different scenes?

@TheoreticallyMedia Жыл бұрын

Yeeeeeah, that gets to be tough. Currently MJ doesn't really do consistent characters. They have been talking about adding it in, hopefully soon. That said, there are a few tricks: try "casting" you actor from a famous actor, and then put them in a particular outfit. Like, for a previous project I used "Young Nicholas Cage in a Brown Leather Jacket" -- It wasn't 100%, but kinda close? You can check out the video here: kzfaq.info/get/bejne/l9Gil8132qy1iWw.html (MJ part is about 3 minutes in)

@nextinstitute7824 Жыл бұрын

Hi there, thanks for great vid! Where can I see the /describe handout? Doesn't let me me download. By the way, didn't you do a vid on camera angles, please link it! Ciao!

@TheoreticallyMedia Жыл бұрын

You should just be able to go to this link and download it. Although you do have to enter a price ($0 is totally fine) and then a link should appear. Let me know if you keep running into problems, just shoot me an email at theoreticallymedia@gmail.com and I'll email it to you. Camera Angles I went into a bit of detail on here: kzfaq.info/get/bejne/mtWVgLZznpyuf3U.html I have a refresh on that coming soon, mostly because I need to finish the prompt guide PDF, and I'll likely roll that into a new video! HOPEFULLY, next week!

@WillDanceAlone2U Жыл бұрын

Great video. In my renderings, I find MJ has a tendency to center all subjects right in the center/upper center. Trying any prompts for it to do otherwise is veeeery frustrating. And I have tried all kinds of tricks to move main subjects off center, but it only works 5% of the time. Have you found a way to place the main subject in a certain area of the image, like you would in a camera viewer, say using the rule of thirds? When/If you ever find a way, will you please share making a video on how to specifically move the subject to a place on the image? It is very frustrating to keep throwing prompts at MidJourney's algorithmwall and hope it will stick somewhere and watch the blurry image turn into 4 more images with subjects in the center. Thanks.

@TheoreticallyMedia Жыл бұрын

Working on some experimentation for you-- I know EXACTLY what you're talking about. The other thing you run into is the "Character with Back to Camera"-- MJ loves that one! ha!

@ominhcuong9538 2 ай бұрын

Hi Tim, i have questions about your theory. Is it still true about the Tokens at this moment (through many updates)?

@aludrenknight1687 Жыл бұрын

It's interesting that the super long 294 tokens prompt provided the image wanted, but supposedly it was capped at 72. I'd imagine if you removed some of the rendering specifications, for example, it wouldn't have the desired result. So what's going on? Is it breaking it down into separate 72 token passes or something? Very interesting.

@TheoreticallyMedia Жыл бұрын

It's weird, and I'm not sure I did a great job explaining it-- but, according to documentation: The bot will attribute 72 tokens across your prompt (no matter how long it is), but then certain words will have more "token power" than others. So, technically, that long prompt has 72 tokens spread across it-- just VERY thinly. To combat that, MJ attributes more "token power" to certain words-- although we don't know what words it chooses. Which, honestly, to me says "MJ ignores words"-- but I guess in the strictest programatic sense, it isn't ignoring the word, but rather attributing zero tokens to it. At some point, I'll do a Token Experiment video, where I take a long prompt and systematically remove words to see what the overall affect is. I'm super fascinated by the whole Token system, and I don't think anyone on KZfaq really talks about it much, mostly because its such a mystery.

@GrisslesWorld Жыл бұрын

Whats the link to that Tokenizer site he was using?

@TheoreticallyMedia Жыл бұрын

novelai.net/tokenizer Sorry, forgot to pop that in the video description!

@user-dp2ny3ov2u Жыл бұрын

Don't forget to put Promt in the video description

@TheoreticallyMedia Жыл бұрын

Wait, which prompt am I putting in? Haha, sorry/- I remember I said I would, I just forgot which one! Too many comments!! Sorry!!

@tr3der Жыл бұрын

is there any type of free MidJourney plan past the free trial?

@voric2893 Жыл бұрын

How are these output generated from the 4 points in the description? is there a way to load these descriptives in?

@Lensman64 Жыл бұрын

2:56 😂

@TheoreticallyMedia Жыл бұрын

Ha’ I’ll never own a real one, but at least I can have a virtual Lens!

@Shanghaimartin Жыл бұрын

5:24 I have seen you pop this box up on several videos but it's not happening with me. How's it done?

@TheoreticallyMedia Жыл бұрын

Ah, that is the Variations button. You’ll see it under the Upscale buttons (U1,2,3,4). Hit the corresponding V button and you can change your prompt to re-output. Be aware though: if you change it too much, it often breaks the image!

@Shanghaimartin Жыл бұрын

@@TheoreticallyMedia Nope. I've hit the V buttons many a time and I've never been given a prompt box before the variation happens :( It just goes ahead and does the variation

@Vulcandream Жыл бұрын

The Indonesian style you mentioned is, to my understanding, pronounced "bateek".

@TheoreticallyMedia Жыл бұрын

Thank you!! That’s Bah-Teek, right?

@Vulcandream Жыл бұрын

@@TheoreticallyMedia that's right.

@Phisynth Жыл бұрын

Interesting video, but FYI, Norman Foster is NOT the architect behind Marina Bay Sands in Singapore, Moshe Sadie is. AI not always telling the truth ;-)

@TheoreticallyMedia Жыл бұрын

Ahhh- THANK YOU! And yup, the age of double fact checking is upon us! That's the 4th lie AI has told me this week!! Well, on the plus side: Now I have TWO architect names to study!! Thanks so much for the correction!

@theheartshines Жыл бұрын

Try getting the prompt with /describe then using that prompt with the reference image!

@TheoreticallyMedia Жыл бұрын

That's a great idea! I'll do a follow up on this soon and incorporate that in! I always love creating those weird loops with Midjourney and other AI systems-- Things can get a little unstable, but it can also lead to some super creative results!

@theheartshines Жыл бұрын

@@TheoreticallyMedia I won the last impossible prompt challenge on Midjourney (table flip). Sometimes the only way to get what you’re looking for is to push the system to intentionally be unstable!

@TheoreticallyMedia Жыл бұрын

@@theheartshines That sounds like solid life advice as well!!

@i_alecio Жыл бұрын

One thing I found interesting with /describe, I got a prompt back and it says that certain word was not allowed. I think it said something about body paint, but the photo was full of texture, leaves, and roots... there was nothing about nudity, who knows

@TheoreticallyMedia Жыл бұрын

That's interesting for sure! Yeah, I know the banned word list has been problematic and they've been working on solutions to make the language model more context aware. At one point, "Cutting" was banned (likely due to self harm triggers), but that also meant you couldn't say "Cutting into a cake" I'm guessing MJ didn't connect the banned word list to /describe, because the language model would melt down trying to dance around alternate ways to rewrite an idea!

@Knight_Industries_2000 Жыл бұрын

@@TheoreticallyMedia "naked" is obviously banned too, so it gave me beautiful back tattoos on skin, but they all end in a collar and short sleeves like the tattoo was the shirt.

@omgchrism Жыл бұрын

so Midjourney sees the Maxtrix... got it!

@TheoreticallyMedia Жыл бұрын

Ha! Did you know that the code in the Matrix is from a Sushi cookbook? The guy that designed the effect scanned in his wife's cookbook and then did a bunch of VFX work-- but yeah, under it all: Sushi Recipes!!

@smartalecc Жыл бұрын

As a 100% newbie to AI, I want to do a small project (a 10 min animation), where the only stipulation is that AI is used for every stage. Can someone correct or improve my rough process below please? This is what I'm planning to do: 1) Script - write with Chat GPT 2) Storyboard - create images with Midjourney 3) Animation - use storyboard images (created with Midjourney) for style + homemade video recordings for motion, and feed into Deep Motion, to create the end animation. 4) Audio - export animation from Deep Motion to Premiere then record character voices on a voice-distorter AI (don't know which one yet), then create final edit on Premiere. -Does someone have a better software idea to input at any of these stages? Many thanks in advance!

@MONTY-YTNOM Жыл бұрын

yet in the 'help' section on MJ it says it doesn't read camera types of lenses to make the picture. Also says 8K doesn't s do anything. Seems these prompts go against it's own rules

@TheoreticallyMedia Жыл бұрын

Yeah. The thing is I think when it says you don’t NEED it, it isn’t saying there aren’t keywords for it. Keywords like Digital Camera or 35mm Panavision vs Leica or Brownie, for sure trigger a difference

@MONTY-YTNOM Жыл бұрын

@@TheoreticallyMedia yer its very confusing :)

@steve_aventura Жыл бұрын

Can you please point in the direction of what a token is. I had not heard it used with respect to midjourney

@TheoreticallyMedia Жыл бұрын

You can check out this link for a brief explainer. But basically, a Token is how a AI Language Model breaks down your input. Like, you write something, and then the AI assigns tokens to your words to "read" them. Here's the link: help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them

@brennakohlhase4194 Жыл бұрын

Bet-eek is how you pronounce batik😊

@TheoreticallyMedia Жыл бұрын

Thank you!! I’m sorry, I’m the worst with pronouncing unfamiliar words, but I’m always happy to announce that I’m probably screwing it up! I’m off to say Bet-eek 20 times and hope it sticks in my noggin!

@aludrenknight1687 Жыл бұрын

thanks!

@VengerSatanis Жыл бұрын

Cha'alt!

@madanglesgaming 8 ай бұрын

The "describe" link you posted in the description is NOT WORKING. Can you please check?

@TheoreticallyMedia 8 ай бұрын

Hey, thanks! I switched links a little while ago and I must have missed this one. Updated on the page, and here it is: theoreticallymedia.gumroad.com/l/describe Thanks again for the heads up!

@spicer41282 Жыл бұрын

Not trying to be nitPicky? Tim. But i got confused when listening instead of watching? So what you described as a BackSlash? \ Is actually called a Forward Slash / (the top is leaning to the right) But I'm sure you meant forward slash / it just came out as back slash ? \ top leaning to the left... Just my 2 cents... Great 👍 Vid BTW!

@TheoreticallyMedia Жыл бұрын

Ha! No worries! Yes, I totally meant to say forward slash! Ugh- I either need more coffee or less coffee apparently! Good catch on the flub though! And thank you!!!

@pluboy444 Жыл бұрын

Midjourney moving fasttttttt

@TheoreticallyMedia Жыл бұрын

Honestly, it's great to see. I was a little critical a few months ago that MJ felt it was slipping. There was a lot happening on the Stable Diffusion side, and MJ was kinda quiet. It's apparent now that silence was them grinding. I think post v5, we're going to be seeing a LOT of innovation!

@pluboy444 Жыл бұрын

@@TheoreticallyMedia agreed😁

@supersupersocco Жыл бұрын

@0:36 why you you prompt for a photo and prompt for CGI and cel shading at the same time? Sign of someone throwing everything at the wall and hoping that helps. These mega prompts make me roll my eyes. You can achieve the same result in 10 words.

@TheoreticallyMedia Жыл бұрын

Oh, that's kind of the point later in the video. To be fair, that one wasn't my prompt, but rather one that I nicked off the MJ community site. I generally agree with you on shorter prompts. If you check out toward the end of the video, I actually showcase via tokens, just how much of that prompt MJ is likely ignoring.

@supersupersocco Жыл бұрын

@@TheoreticallyMedia ah nice, perhaps could have addressed tokens right after that mega-prompt.

@TheoreticallyMedia Жыл бұрын

@@supersupersocco Haha, a TWIST ending worthy of M. Night!

@geolitz2665 11 ай бұрын

Does anyone know who prompted the thumbnail for this video?

@TheoreticallyMedia 11 ай бұрын

Gotcha on the Image video!

@geolitz2665 11 ай бұрын

@@TheoreticallyMedia What?

@TheoreticallyMedia 11 ай бұрын

@@geolitz2665 I think you asked this question twice in different videos. I answered you in the other one. Did you get it?

@geolitz2665 11 ай бұрын

@@TheoreticallyMedia I did. Thank you.

@joesaad2166 Жыл бұрын

Hello marina bay sands architect is Moshe Safdie

@TheoreticallyMedia Жыл бұрын

Thank you! I guess Midjourney needs to study architects better! I’m going to go look at more Moshe buildings now. I think the Bay Sands is such a cool structure!

@joesaad2166 Жыл бұрын

Yes indeed .but all what is happening with the ai is amazing.thanks for the tutorial

@AmberCox85 Жыл бұрын

I tested this on my own work and not even close • impossible to get the prompt for any image, especially if you use your own image for input

@TheoreticallyMedia Жыл бұрын

How's so? I got the Waterfall prompts?

@AmberCox85 Жыл бұрын

@@TheoreticallyMedia it noticed things like colors or in your case a waterfall, but was way off from my design that I was asking for it to prompt describe. I’m sure it will develop itself. However, it did generate some cool prompt ideas

@rodee1671 Жыл бұрын

Buh'-teek = batik.

@TheoreticallyMedia Жыл бұрын

Thank you!! I really should start checking with Google Translate for these videos!

@AvaTellsTales Жыл бұрын

Looking for a channel that combines AI technology with stunning artistry? Look no further than Ava Tells Tales! Our short stories are crafted with ChatGPT, and brought to life with beautiful visuals from Midjourney and animations from D-ID!

@TheoreticallyMedia Жыл бұрын

I watched "The Hunt" and Mark's Story -- That's GREAT! Loved the whole thing! I played around a bit with some horror shorts using MJ (they're on the channel)-- the idea being that it would be a series of stories told through the camera roll of a phone. Lots of fun to put together! But, the whole Shorts/Long Form on one channel really screws up my analytics. At some point I should just do them on another channel. So, just a thought on yours: I presume you're using ElevenLabs for the voice? I think if you could use a human voice for the read, it would really pop. Like, if you could cast an Ava. While I do feel Ava's current monotone/flat read works very well for her stories, to me, a real person read would kick things up a notch. Plus, viewers would have to play that game of "Is it AI, or is it real?" which would add a whole new dimension to it! That said, fantastic work! Keep it up!

@AvaTellsTales Жыл бұрын

@@TheoreticallyMedia Wow, thank you so much for your comment and feedback. Normally, the AI writes everything about this account, including my first comment on the video, but I wanted to reply personally :) Actually, I use D-ID for voice and animation. I tried all the tones and moods in D-ID and thought this was the most appropriate tone for the character. Regarding the voice acting by a real person, I can say that the purpose of this project is to explore the results of creating all elements with AI. Of course, a real person handles the prompts and video editing, but I wanted to see what the outcome would be if AI did everything else. These were the first videos I attempted, but for future videos, I plan to make the stories more impactful, add a little more three-dimensionality to the visuals, and enhance the immersion. Thanks again for your feedback!

@TheoreticallyMedia Жыл бұрын

@@AvaTellsTales Oh, I 100% the "entirely generated by AI" angle! In that case, HIGH recommend to try out ElevenLabs for Ava's voice. There's a video I did on GPT4 vs Elon where I used D-ID and Eleven Labs to read a ChatGPT response of the Open Letter calling for the pause of AI. I cloned Benedict Cumberbatch's voice, and the results are super impressive. But yeah, just a thought! Overall, I really love what you're doing and can't wait to see how it all progresses!

@AvaTellsTales Жыл бұрын

@@TheoreticallyMedia Thank you, I'll definitely check it out!

@MFJLabs Жыл бұрын

Background music is too loud

@TheoreticallyMedia Жыл бұрын

Ha! Pretty common complaint on my channel. Working on it. Sorry, it’s a lot like how it is in my house, everyone yelling: “Turn your music down!!” I’ll keep an eye on the VU meters next video!

@user-tx7yv9ue7m Жыл бұрын

hi man, cool work! Please help me create one scene, I spent all my hours to generate them, but I didn't succeed( I would be very glad if you could help me create such a scene. So, scene 1: what would a child see when he is abducted, that is, the view from the child's face

@user-tx7yv9ue7m Жыл бұрын

or, what would a person see who fell from the roof of a tall building (also a view from the face of a falling person)

@TheoreticallyMedia Жыл бұрын

I would try POV shot (Point of View)-- that might get you there. Ha, should I ask what this project is for? It does sound pretty dark!

@user-tx7yv9ue7m Жыл бұрын

@@TheoreticallyMedia thank you man👍 not really, I want this kind of project to call for the protection of the lives of children and people) BUT the “prompt with POV picture” shows on behalf of the kidnapper (in the scene with the abduction)

@user-tx7yv9ue7m Жыл бұрын

@@TheoreticallyMedia sorry man, but i was tormented, prompt pov didn't work...you would make me very happy if you try to create these scenes on Midjourney

@twitterglobalarmy Жыл бұрын

Midjourney is cool but users must pay to utilize the site after a brief period of free use. I've also experimented with Blue Willow and I'm really amazed to use it

@f4ust85 Жыл бұрын

So basically this will lead to even more chaos, confusion and gibberish obscuring what is ACTUALLY on the picture with buzzwords and pseudo-helpful formulas and clichés. As if it wasnt enough how much nonsense and totally contradictory stuff "proompters" put into their prompts thinking they are smart. Like in the first example: 50mm lens, wide angle, Unreal Engine, RayTracing, OpenGL, Tone Mapping...? UGH! These things are pretty much totally negating one another, let alone realistic results.

@TheoreticallyMedia Жыл бұрын

To be clear, that wasn't actually my prompt. I just nabbed it off the community feed. Those mega long Prompt-Gore/Splatter Prompt kind of wall of text is something that I generally preach against. My point in this video is that /Describe gives us the buzzwords that Midjouney actively knows. That's actually super important. Towards the end of this video, I also discuss Tokens, and MJ's Token limitation-- which, basically proves that more than 3/4 of that mega long prompt is actually being ignored.