Ione erklärt die Welt von #Uniworld von Ione - Dailymotion

Ione

Was ist eigentlich dieses #Uniworld? Wie funktioniert ein Pen & Paper System? Was ist das überhaupt? Und was ist in den nächsten Jahrzehnten eigentlich passiert? Ione erklärt dir die Hintergründe zur #Welt von #Uniworld, was du brauchst, um es zu spielen und was man hier so alles machen kann.

Transcript

00:00:02Hi all, and welcome to my tutorial for AI Music Video Creation.

00:00:09I did a live stream yesterday for Bandcamp.

00:00:16Unfortunately, my voice wasn't recorded, so I'm basically recording now.

00:00:22I'm recording my voice now above the video, so it may be a bit clunky

00:00:28at times. Right now I'm still setting things up, and yeah, and let's get started.

00:00:40So the first thing I showed was that I split up my original music track into

00:00:48several parts and created an image for Ioni that I want to use for the

00:00:57music video. My idea was to have Ioni do a simple walking animation while lip-syncing

00:01:06with a song. And yes, for that I split up the song into multiple parts, basically all

00:01:20around a little bit less than 40 seconds in length. The reason for that is that I'm

00:01:28going to use LTX 2.5 as a video model. And the LTX has a, or my PC is capable

00:01:43of around

00:01:43doing 20 seconds of video in one go. So basically, if I create a video, it has

00:01:53to be split up in multiple parts. And for that I'm going, or for the video creation, I'm using

00:02:03Maestro via Pinocchio. Pinocchio is basically a one-click installer tool. And Maestro is more or less

00:02:13for a solution where you can create videos, images, and music locally on your PC.

00:02:22And if you don't want to hang around with installing all of the apps and dependencies

00:02:28yourself, you can use Pinocchio. Just download it, install it, then discover what there are hundreds

00:02:39of available AI tools that you can use. And yeah, the one I'm using here is Maestro.

00:02:49It's more or less, well, the successor of 12GP. I mean, 12GP is still out there. But if you look

00:02:59at Maestro, that's well, way better. And well, also the original 12GP GUI is included as well.

00:03:07So that's why I switched to Maestro. For the more I had some trouble with 12GP lately, where

00:03:16it had, when I created a video with Altix, there was always subtitles in the video, but not,

00:03:23well, let's say real subtitles. Instead, there were, well, fantasy text, let's say like this,

00:03:32there's broken text, nothing to do with the, well, with my original video and text.

00:03:41So I have no idea where that came from. I even installed a LoRa, which is, which is exactly

00:03:47for that to not have the subtitles, but didn't work. So that's why I switched to Maestro's.

00:03:56No, no longer had any problems. I don't know. So I'm giving a short direction of what you

00:04:05can do here, Pinocchio. There are hundreds of available tools you can install, all

00:04:10with basically one click. So no longer having to try to get something to run. I think most

00:04:19of the stuff even works on Mac and not only on PCs. So I know that Macs have been always

00:04:27a little bit problematic when it comes to AI tools. So you might even give that a try as

00:04:34a Mac user. Okay. So Maestro has booted up. That's the UI of Maestro.

00:04:44Basically it's, right now I'm in the view where you can see everything. Then there's images, videos,

00:04:53and audio. So far I haven't used it to create audio, so it's empty.

00:05:01Yeah. There is a director mode, which you can use to create, basically it's an agent, which

00:05:09you can use to create videos. I've tried it a few times, but well, I wasn't really impressed

00:05:17with the quality of it. The main issue was that it wasn't capable of creating good images.

00:05:28I don't know if that has to do with the image model. I'm using quen image edit 2511, which

00:05:35is basically the best image model you can have locally. So yeah, I'm mainly using it for

00:05:45the normal studio mode to create videos. I've got ATX 2.5 and Minimax H3 installed.

00:05:55Minimax H3 for testing purposes, because basically it's not, you're not allowed to use it in the

00:06:03European Union where I live. So yeah, but Minimax is still a very, very good model.

00:06:15And it can also do lip-syncing, so you can provide your own audio and it does lip-syncing to

00:06:20that.

00:06:22But since for that video, I only want to have more or less simple walking animation for

00:06:30Ioni, I'm going to use ATX 2.5, especially because my hardware can do longer videos with

00:06:41ATX than with Minimax. So Minimax is around, I think, seven seconds or so. And that's what my

00:06:48hardware is capable of. And for ATX, I can also, I think I've tried 30 seconds, but the default

00:06:57video length is 20 in one go. So now I'm setting up my video creation. So that's an image I

00:07:12created

00:07:12of Ioni steampunk outfit. So the song is about steampunk world. And so going to the studio mode.

00:07:27And there you can see, you can select the model down below and describe what you want to see

00:07:38in the scene. It puts frames and you can put in a soundtrack, which I'm going to use here.

00:07:50So I'm using that image as a start frame. And then I'm using that first part of song as a

00:07:58soundtrack.

00:07:59Please note that the video duration is now 20.9 seconds, although the audio is 38.

00:08:04As you can see, that is the reason why I haven't checked that box. And now I'm putting in the

00:08:17audio again.

00:08:18And you will see that the video length above the duration went to 38 seconds. And also important there,

00:08:26it says two windows. That means two sliding windows. I'm going to explain these in a minute.

00:08:37So basically what the sliding window does, as you can see, it's 20.9 seconds here.

00:08:46Sliding window, the total video is 38 seconds. But my PC will create a video of 29 seconds first.

00:08:55So let's say almost a half of the video. And then I will create a second video using for the

00:09:03remaining half.

00:09:05The problem of sliding windows is, as you can see, I've created a full video, three minutes, using that.

00:09:16And have a look now. You might have seen a small distortion right now. So I'm thinking I'm going, yes.

00:09:25So that's the first sliding window. And now you've seen it. The color changed.

00:09:31And that's how it looks in the later sliding windows. So yeah, there you get quality loss and consistency loss.

00:09:42That's why one sliding windows might be okay. Now I'm just showing the second. So even there, two might be

00:09:52okay.

00:09:52But never go above that. Solution for that is simply create shorter videos and then extract the last frame of

00:10:05your original video.

00:10:06And start from starting new. Yes, right. Now I need a prompt as well for the video. So, um,

00:10:36for that I've prepared, um, uh, something. Let's see. Uh, yeah, as I said, um, it's, I'm speaking above a

00:10:46recording. So, uh,

00:10:49I'm not really sure what I was doing at the time. Um, ah, right. I'm now, uh, I'm using the

00:10:57same image as start and end frame. So basically the video starts or stops where it started so that I

00:11:05can basically create a loop. So that was my idea above that, uh, for that. And the general idea for

00:11:13my music video here, um, for, um,

00:11:16Tears of Breast and Bone is the Hive have Ioni singing and, um,

00:11:23and, um, so make her transparent, uh, like 70% or so, so that you can only see her a

00:11:29little bit and then have a video playing above her or at times blend her in completely.

00:11:38So that's why I don't need, um, yeah, uh, really good, um, um, uh, really awesome video for her, just

00:11:49her singing. Um, so, um, yeah, I was talking about frames earlier. Um, how to, there's a tool I'm using,

00:11:59by this frame extractor, um,

00:12:03um, that allows you to extract frames from a video.

00:12:07Um, so, um,

00:12:09um, I always set it to 30 frames upload the video and then click the extract button.

00:12:14Um,

00:12:15and then you can see every image of that video,

00:12:19um, here, that one is from Monica from, uh,

00:12:24Doki Doki Literal Club, very, very good game.

00:12:28Yeah, you can download the image.

00:12:31So that goes, I downloaded the last frame and there it is.

00:12:35And so I could use that for, yeah,

00:12:40for the video as a first frame for the next video,

00:12:46basically, if I wanted to.

00:12:48So I use that quite often.

00:13:00So I'm going back to LTEX.

00:13:05So now, as you can see the script,

00:13:09the video idea box is empty.

00:13:14What I did, I'm using Google AI Studio

00:13:17and have basically, I think I'm, yes.

00:13:22I've uploaded several prompting guides,

00:13:26for example, C-Dense 2, C-Dense 2.5,

00:13:30LTEX 2.5, Minimax 83.

00:13:34And I gave Google Gemini the link

00:13:40and description to the prompting guide.

00:13:42So I have saved them on my PC.

00:13:46There are several of the guides.

00:13:48So for example, that's one for Minimax H3 from Icora AI.

00:13:55Then there's two more for Minimax H3 directly

00:14:00from the official GitHub,

00:14:02from the official Hugging Face site.

00:14:11So basically, with that, feed the prompting guides

00:14:14to your AI of choice.

00:14:17And then, yes, you can tell them to make your video.

00:14:26So why is it important?

00:14:28For example, that is, I created that video,

00:14:31it was Minimax H3.

00:14:33And this is how a prompt for Minimax H3 looks like.

00:14:39And at least for me, no human can write that anymore.

00:14:42So you need to have an AI companion

00:14:46to write these prompts for you.

00:14:48Otherwise, you won't be able to create videos.

00:14:52So that's why these prompting guides are so important.

00:15:02So, and now I will run to a little problem right now, later,

00:15:10with that chat.

00:15:14Google AI Studio is a little bit particular.

00:15:17When you delete an image or so, you upload it here earlier,

00:15:21which I apparently did.

00:15:23Your chat doesn't work anymore.

00:15:26So I will find that out in a minute.

00:15:30For now, I uploaded the image here.

00:15:34And then, as you can see now, I'm telling Gemini

00:15:41that it should write a prompt for LTX 2.5.

00:15:44This is my chat with the LTX 2.5 prompting guide.

00:15:49I will create a new one in a minute.

00:15:52Of course, this one doesn't work anymore.

00:15:57And yes, I'm telling Gemini that it should write a prompt for me

00:16:03using the prompting guide that I feed it earlier.

00:16:07The Ellen woman is slowly walking while singing.

00:16:12I also tell it that I will provide the audio to the video generation tool.

00:16:20And I think right now I'm going through the, looking for the subtitles.

00:16:28I believe.

00:16:31Two monitors and unfortunately in the recording,

00:16:34I can't see what I did on the second monitor.

00:16:41But I think I was opening my subtitles with the,

00:16:48ah, that's what I did.

00:16:50So I split, as I said earlier,

00:16:53I split the audio into multiple parts.

00:16:56So to six parts in total.

00:16:58And to have them below 40 seconds.

00:17:01So that means, ah, be the, all of the videos are,

00:17:07uh, smaller or only use maximum of two sliding windows.

00:17:12I gave, um,

00:17:14Gemini the, uh, lyrics of my song.

00:17:18And told it to create an prompt for,

00:17:21or create a command for FFmpeg.

00:17:23Which is a tool for,

00:17:25which we also can,

00:17:27which can do lots of things.

00:17:28including splitting up audio files.

00:17:32And, uh, as you can see for the first, um,

00:17:37uh, the first, uh, split, uh,

00:17:39includes, uh,

00:17:40first eight lines of the song.

00:17:43So I'm right now copying the lines.

00:17:47And, yeah,

00:17:48we moved, uh, the time codes.

00:17:50So that I have only the text that is sung in there.

00:17:54Uh, going back to our text.

00:17:56And now, uh, I'm telling, uh, uh,

00:18:00uh,

00:18:02my Gemini and I tell, um,

00:18:05Gemini that it should sing,

00:18:07uh,

00:18:08create a prompt where she's singing.

00:18:10And what I said earlier,

00:18:12this thread is broken.

00:18:13As you can see,

00:18:14internal error occurred.

00:18:15That basically means I probably deleted one of the images

00:18:19I uploaded, uh,

00:18:21there so that it doesn't like.

00:18:23So what I simply doing,

00:18:25I'm creating a new one.

00:18:27Um,

00:18:28I found out that I don't have the prompting guide for

00:18:32LTX 2.5, um,

00:18:34saved here.

00:18:35So in a minute,

00:18:36I'm going back to my old chat.

00:18:40Um,

00:18:41there it is.

00:18:42Scroll up, uh,

00:18:43to the top and there's the prompting guide.

00:18:45So,

00:18:46but this time I made sure to save it up as well.

00:18:50So you can see,

00:18:51I'm creating a new text file,

00:18:52LTX 2.5.

00:18:55And then, uh,

00:18:57on the other screen,

00:18:58I saved it there.

00:19:00And so I'm telling now Gemini in my new chat,

00:19:04that it should be,

00:19:05uh,

00:19:05make itself comfortable with the prompting guide for LTX 2.5.

00:19:15Uh,

00:19:26enabling UL context is important.

00:19:28Otherwise it can't open URLs.

00:19:31And now it's reading with website.

00:19:33And it's telling me what it found out.

00:19:37And now I can go back to my first chat,

00:19:42where I wrote the prompt earlier.

00:19:45And,

00:19:45um,

00:19:47yes.

00:19:51So.

00:20:00Okay.

00:20:03Ah,

00:20:04there.

00:20:07I'm not sure what I was doing right now.

00:20:10Ah,

00:20:10right.

00:20:10I also,

00:20:12um,

00:20:13uploaded the image there.

00:20:14That's important.

00:20:16Um,

00:20:16so now I'm copying the prompt I already wrote.

00:20:20Um,

00:20:20close the tab because I don't need it anymore.

00:20:23And,

00:20:24um,

00:20:26yeah,

00:20:27then,

00:20:28uh,

00:20:31I also provided the audio.

00:20:33just to give as much detail as possible.

00:20:37And what did that?

00:20:38I also attached the audio file,

00:20:40uh,

00:20:41here.

00:20:43And then I hit the run button.

00:20:50and there we go.

00:20:52So that's the prompt.

00:20:55It gave me to,

00:20:57which I can just use,

00:20:59um,

00:21:00to copy,

00:21:01uh,

00:21:02to,

00:21:02uh,

00:21:03to,

00:21:03my scroll.

00:21:05So basically,

00:21:06uh,

00:21:07uh,

00:21:07the,

00:21:08uh,

00:21:08the,

00:21:08uh,

00:21:09wire and brass goggles.

00:21:11Walk slowly across snow-covered classical stone ruins.

00:21:16And there we go.

00:21:19So,

00:21:20I'm checking the settings.

00:21:21Um,

00:21:23I want it to be 9 to 16 format.

00:21:26So I'm selecting the format sliding window.

00:21:30This 20.9 seconds.

00:21:32And,

00:21:33uh,

00:21:33what I wanted,

00:21:351080p.

00:21:36And then I hit the generate button.

00:21:41So it's,

00:21:42now,

00:21:43basically what it does now,

00:21:44it's creating,

00:21:45um,

00:21:45the sliding windows and it will create,

00:21:49um,

00:21:49the videos.

00:21:51Um,

00:21:52so,

00:21:53yeah,

00:21:5438 videos.

00:21:54I think it will take around 50 minutes or so,

00:21:57I believe.

00:21:59Um,

00:22:01that's really,

00:22:02really nice.

00:22:03Yeah.

00:22:06Um,

00:22:07ah,

00:22:07someone was asking about,

00:22:09my hardware specs.

00:22:11So,

00:22:13basically I have,

00:22:15uh,

00:22:17uh,

00:22:17I think,

00:22:18uh,

00:22:18what was it?

00:22:19right.

00:22:20Um,

00:22:20I nine 49,

00:22:22uh,

00:22:2214,900 K is my CPU.

00:22:25I've got,

00:22:26uh,

00:22:2664 gigabytes of Ram.

00:22:28And,

00:22:28uh,

00:22:29RTX 50,

00:22:3070 TI graphics card with 16 gigabyte VRAM.

00:22:34Um,

00:22:36yeah.

00:22:36So,

00:22:37I think,

00:22:39it's 2.5 might work with 12 gigabyte.

00:22:43I believe.

00:22:44I'm not sure.

00:22:45Um,

00:22:47I'm using a special model,

00:22:50which is optimized for,

00:22:52uh,

00:22:52RTX 50,

00:22:53uh,

00:22:54for the RTX 50 series.

00:22:56Um,

00:22:57which is quite nice because it's even smaller than the original model.

00:23:03Um,

00:23:06yes.

00:23:07Um,

00:23:09um,

00:23:10um,

00:23:10yeah,

00:23:11and yeah,

00:23:12this,

00:23:12what I just,

00:23:13uh,

00:23:13did for the first,

00:23:14um,

00:23:15for the first,

00:23:16um,

00:23:1738 seconds.

00:23:18I will do it as well for the other parts of the music.

00:23:23Um,

00:23:24um,

00:23:24I had DaVinci resolve open my video editor,

00:23:27uh,

00:23:27just closed it in case,

00:23:29uh,

00:23:30resource,

00:23:30uh,

00:23:32might,

00:23:33might've used too many resources.

00:23:35Um,

00:23:37so let's see.

00:23:46um,

00:23:47yes,

00:23:48uh,

00:23:53closed, uh,

00:23:54DaVinci resolve in case it used too much of my system resources,

00:23:59so that the operation can run,

00:24:02uh,

00:24:02fine.

00:24:07then,

00:24:07uh,

00:24:08while that was running,

00:24:09I went over to,

00:24:11uh,

00:24:13uh,

00:24:13Meta AI.

00:24:16um,

00:24:17and so,

00:24:18uh,

00:24:18Meta,

00:24:19um,

00:24:20is right now free for,

00:24:21uh,

00:24:21image,

00:24:22uh,

00:24:22for example.

00:24:23So it's a fairly,

00:24:25uh,

00:24:26they introduced a new image creation model fairly recently.

00:24:31And,

00:24:32um,

00:24:33what I'm telling Meta is that I,

00:24:36that it should create images,

00:24:39um,

00:24:40for my music video,

00:24:41because I want to have Ioni,

00:24:44um,

00:24:45walking,

00:24:46um,

00:24:47around and,

00:24:47uh,

00:24:48want to,

00:24:48uh,

00:24:49walk and,

00:24:50uh,

00:24:50in front of her,

00:24:51I want,

00:24:53um,

00:24:54well,

00:24:54basically playing,

00:24:56a real music video,

00:24:58or a real video,

00:24:59um,

00:25:02for that,

00:25:02uh,

00:25:03Meta and also a movie maker,

00:25:05provided by Zelda,

00:25:08um,

00:25:09which is also a really,

00:25:10really good tool.

00:25:11So you can provide,

00:25:12um,

00:25:13story and style for that.

00:25:14You can select the number of frames,

00:25:18uh,

00:25:18remember a number of images you want to create.

00:25:22Um,

00:25:23you can select the,

00:25:24uh,

00:25:25uh,

00:25:26the,

00:25:27uh,

00:25:28uh,

00:25:30uh,

00:25:30resolution you want to have.

00:25:31Um,

00:25:32and the,

00:25:33uh,

00:25:33so for example,

00:25:34it's right now 16.9,

00:25:3616.9,

00:25:3616 to nine.

00:25:39Aspect ratio,

00:25:40not solution,

00:25:41not resolution,

00:25:42the aspect ratio.

00:25:43Um,

00:25:44you can give it,

00:25:45um,

00:25:46location.

00:25:47So hintergrund,

00:25:48Landfaust build,

00:25:49um,

00:25:50the box,

00:25:51uh,

00:25:52the,

00:25:53uh,

00:25:54drop down below where you can upload an image.

00:25:57And what I am doing,

00:25:59I gave,

00:26:01uh,

00:26:01give the same prompt to both meta and the movie maker.

00:26:06And,

00:26:07uh,

00:26:08then,

00:26:08um,

00:26:10ask them to create images for me.

00:26:13So basically I want realistic looking images.

00:26:17And then,

00:26:17um,

00:26:18yeah,

00:26:19steampunk world.

00:26:19And where the automata,

00:26:22uh,

00:26:22automatons have been,

00:26:24uh,

00:26:24or have received a soul.

00:26:27And yeah,

00:26:28they,

00:26:28um,

00:26:29uh,

00:26:29well,

00:26:30now self-conscious and they are rising up against the former human masters.

00:26:36So,

00:26:36and that resulted in conflict.

00:26:38So that's basically,

00:26:40um,

00:26:41what that song is about.

00:26:44And yeah,

00:26:45I'm also telling it that I have a song for this.

00:26:49and,

00:26:49uh,

00:26:50including lyrics.

00:26:52Then,

00:26:54I've,

00:26:55uh,

00:26:57tried to upload the song to meta,

00:26:59but unfortunately,

00:26:59meta doesn't,

00:27:01uh,

00:27:01support audio yet.

00:27:03Okay.

00:27:04I don't know.

00:27:04I'm not over,

00:27:05I don't know what to get,

00:27:06but it doesn't support audio,

00:27:08but at least I gave it the complete,

00:27:10uh,

00:27:10lyrics in ASAT,

00:27:11SRT format.

00:27:16And,

00:27:17um,

00:27:18yes.

00:27:20Um,

00:27:23not only to create,

00:27:24um,

00:27:28uh,

00:27:28yeah,

00:27:29images for,

00:27:30um,

00:27:32uh,

00:27:33to the lyrics.

00:27:40Unfortunately,

00:27:42also my,

00:27:42uh,

00:27:43video recording broke down,

00:27:44um,

00:27:45at,

00:27:46uh,

00:27:46that time.

00:27:47Um,

00:27:48so I'm recording it,

00:27:49again,

00:27:50uh,

00:27:51that part.

00:27:52Um,

00:27:53so I worked,

00:27:54uh,

00:27:55uh,

00:27:56last,

00:27:56um,

00:27:57where I gave it a prompt,

00:27:59um,

00:27:59to create,

00:28:01uh,

00:28:01images,

00:28:03um,

00:28:04and gave it the lyrics.

00:28:06Um,

00:28:07what it then did,

00:28:09um,

00:28:09that took,

00:28:10I think around an hour or so.

00:28:12Uh,

00:28:13what it created,

00:28:14uh,

00:28:14was,

00:28:15um,

00:28:15this.

00:28:18So complete,

00:28:19uh,

00:28:20storyboard.

00:28:21Um,

00:28:22we had,

00:28:22I had some problem with understanding how it works,

00:28:24because of that is.

00:28:26Image one and that is image 25.

00:28:29So usually,

00:28:30I mean,

00:28:30we are used to reading left to right.

00:28:33No,

00:28:33it's,

00:28:34uh,

00:28:34as you can see,

00:28:35one,

00:28:35two,

00:28:36three,

00:28:36it goes down to 24 and then starts above here with 25.

00:28:40Um,

00:28:41but that's what it created.

00:28:43So complete storyboard with,

00:28:46uh,

00:28:47prologue act one.

00:28:48And so it's great.

00:28:50Basically an app,

00:28:51um,

00:28:52took an hour,

00:28:52but,

00:28:53I mean,

00:28:54you can see the lyrics up here.

00:28:58Um,

00:28:59and that's for that image is created.

00:29:02And,

00:29:03uh,

00:29:03the next lyrics.

00:29:05Um,

00:29:07and,

00:29:07uh,

00:29:08yeah,

00:29:08so basically what it did,

00:29:10um,

00:29:11it created,

00:29:12uh,

00:29:12images for these lyrics.

00:29:15And,

00:29:16um,

00:29:16yes,

00:29:17I also,

00:29:19um,

00:29:20yeah,

00:29:20used the movie maker.

00:29:22Um,

00:29:22unfortunately you can't see,

00:29:24what you did in the past here.

00:29:27so,

00:29:28um,

00:29:30uh,

00:29:30opening,

00:29:31uh,

00:29:32let's see,

00:29:33where do I have it?

00:29:34Um,

00:29:37let me quickly open,

00:29:38um,

00:29:43um,

00:29:44it created a zip file.

00:29:46So you can download,

00:29:48everything it does,

00:29:48um,

00:29:50after you click on storyboard,

00:29:51um,

00:29:53or maybe a little bit more to that tool.

00:29:55Um,

00:29:56usually you put in character or multiple characters,

00:29:59and these are present in every image.

00:30:01So if you have a main character or want to create a story or so,

00:30:04that's good.

00:30:06Um,

00:30:07but you can also write in the prompt that you don't want to have any character.

00:30:11You have to edit character,

00:30:14but you can,

00:30:14uh,

00:30:15so,

00:30:16uh,

00:30:17can also do something like this and then have a character,

00:30:20which,

00:30:20which is no correct on the divide,

00:30:22uh,

00:30:23uh,

00:30:23up above here.

00:30:24You don't want any main characters or so.

00:30:28and then it created,

00:30:29I think I've selected 40 images.

00:30:31And then it creates,

00:30:33uh,

00:30:3340 images,

00:30:34um,

00:30:34using nano banana one.

00:30:36That's the only downside of the tool.

00:30:37Um,

00:30:38but it's three.

00:30:40And these are the images it created.

00:30:43Uh,

00:30:44it also creates,

00:30:45um,

00:30:46a small storyboard document.

00:30:49So these are the lyrics I provided,

00:30:53basically.

00:30:54And this is what happens in that scene.

00:30:57So a short description of the scene of the image.

00:31:02and then what I did,

00:31:04uh,

00:31:04I selected this image because,

00:31:07um,

00:31:08the according,

00:31:09uh,

00:31:09accompanying is,

00:31:11uh,

00:31:12so,

00:31:13uh,

00:31:13where is it?

00:31:14Yeah,

00:31:14the old copper world gave machines the spark to feel.

00:31:17So very clear.

00:31:19Um,

00:31:19yeah,

00:31:20he gave it,

00:31:21um,

00:31:23yeah,

00:31:23he gave them the soul and I liked the image.

00:31:26So that's then called copper world.

00:31:28So scientist,

00:31:31or,

00:31:31uh,

00:31:32engineer,

00:31:33and who gave the automaton,

00:31:35um,

00:31:36well,

00:31:36the spark to feel.

00:31:40And yeah,

00:31:41using,

00:31:41I wanted to use that image.

00:31:43and now we can also,

00:31:47um,

00:31:49continue,

00:31:50wait,

00:31:50continue with that one.

00:31:53So,

00:31:54then I went to,

00:31:57uh,

00:31:58KZ.AI.

00:31:59So my main platform where I create,

00:32:02um,

00:32:03videos,

00:32:04um,

00:32:04aside from local generation.

00:32:06Um,

00:32:10uh,

00:32:11yes.

00:32:12So,

00:32:12uh,

00:32:16um,

00:32:17not sure.

00:32:18I'm pretty sure I'm doing something on my other monitor again.

00:32:21So let's skip a few seconds ahead.

00:32:25Uh,

00:32:26let's go there.

00:32:28Okay.

00:32:30So then I have a gem or I have a chat on normal Gemini,

00:32:35which,

00:32:37well,

00:32:37I use for,

00:32:38let's say other models,

00:32:39which don't have,

00:32:41um,

00:32:41specific prompting guide.

00:32:43And,

00:32:45uh,

00:32:45yes.

00:32:46So what I did here is,

00:32:48um,

00:32:49I,

00:32:50uh,

00:32:51I'm using,

00:32:52uh,

00:32:53yeah,

00:32:53this,

00:32:54this chat,

00:32:55um,

00:32:56to create a prompt for Cdense 1.5.

00:33:04or six second video.

00:33:04So let's see what I'm doing.

00:33:06Um,

00:33:10well,

00:33:10there,

00:33:10I'm using that image.

00:33:14And upload it there.

00:33:19And then I'm telling Gemini to create a prompt for,

00:33:23uh,

00:33:23Cdense 1.5 for six second video.

00:33:32you see danced.

00:33:34Okay.

00:33:35And,

00:33:35uh,

00:33:36yeah,

00:33:36to insert,

00:33:37um,

00:33:39yeah,

00:33:40the,

00:33:40the bottle,

00:33:41or how to explain,

00:33:45uh,

00:33:48glowing bottle,

00:33:50I believe.

00:33:50Yeah.

00:33:51Into the robot.

00:33:53and then robot,

00:33:55uh,

00:33:55yeah,

00:33:55comes to life.

00:34:00Hmm.

00:34:01Okay.

00:34:01I'm seeing spelling mistakes,

00:34:03but,

00:34:03uh,

00:34:03well,

00:34:04good thing.

00:34:04It's an AI,

00:34:05uh,

00:34:06LRM.

00:34:06Um,

00:34:07they know what I want to do.

00:34:09Uh,

00:34:09so life should have been with an F and see that.

00:34:12Uh,

00:34:13yeah,

00:34:14but anyway,

00:34:15um,

00:34:16so you can see,

00:34:17uh,

00:34:18it correctly,

00:34:19uh,

00:34:20notice.

00:34:21Okay.

00:34:21You,

00:34:21you probably want to use Cdense 1.5.

00:34:23And then that's the,

00:34:25uh,

00:34:26prompted,

00:34:26uh,

00:34:27wrote me.

00:34:27So the,

00:34:29last file is inserted into the robot.

00:34:32And then,

00:34:33yeah,

00:34:33blue energy searches for the,

00:34:35the robot.

00:34:36And it basically comes to life.

00:34:39And,

00:34:40uh,

00:34:40Moose wakes up.

00:34:44Uh,

00:34:45I think I forgot.

00:34:46Uh,

00:34:46I've been,

00:34:47I'm uploading the image here to az.i.

00:34:50Um,

00:34:52now copying the prompt as well.

00:34:56And so,

00:34:57um,

00:34:58this is not needed,

00:35:00um,

00:35:00for Cdense 1.5.

00:35:04and so I want a six second video.

00:35:07I don't need any audio.

00:35:09Um,

00:35:101080p.

00:35:10Also,

00:35:11I have my own audio for the song.

00:35:13And then I basically just hit the,

00:35:16um,

00:35:17create button.

00:35:19And then that takes some time,

00:35:22um,

00:35:24for the generation to complete.

00:35:32So,

00:35:33yeah,

00:35:33as you can see,

00:35:34uh,

00:35:35at that point in time,

00:35:36Meta was still working,

00:35:39um,

00:35:40on,

00:35:41uh,

00:35:41yeah,

00:35:41on the,

00:35:42uh,

00:35:43images.

00:35:44Um,

00:35:45I've shown you the results before.

00:35:47Um,

00:35:48so it took,

00:35:48I think really took like an hour or so,

00:35:51for it to finish.

00:35:54Um,

00:35:55yeah,

00:35:56but,

00:35:56uh,

00:35:56I think the results were pretty good.

00:35:59In the end,

00:36:00but,

00:36:01uh,

00:36:01maybe I should have told it to,

00:36:04uh,

00:36:06yeah,

00:36:07just,

00:36:07uh,

00:36:08create,

00:36:08uh,

00:36:09uh,

00:36:11images without an app or so.

00:36:13And my local generation also completed.

00:36:18See,

00:36:19so the first video is 86 megabyte.

00:36:21Then the complete one's nearly 160.

00:36:24So basic what it did,

00:36:26it created,

00:36:27ah,

00:36:27let's see it.

00:36:50So I was,

00:36:52pausing quickly so that you can see,

00:36:55um,

00:36:56that,

00:36:56uh,

00:36:58yeah,

00:36:58a small,

00:36:59small distortion,

00:37:00um,

00:37:01when it came to the next sliding window.

00:37:04So,

00:37:04but it's still okay.

00:37:05So as I said,

00:37:06one sliding window,

00:37:08um,

00:37:09probably,

00:37:09or it's okay,

00:37:10I think.

00:37:12And,

00:37:12um,

00:37:13then someone was asking,

00:37:15oh,

00:37:15that has lip sync.

00:37:17Uh,

00:37:17okay.

00:37:17then I said,

00:37:18yep,

00:37:18because I provided the audio.

00:37:22Um,

00:37:25yeah,

00:37:26the audio file,

00:37:27which is 38 seconds.

00:37:28And,

00:37:29um,

00:37:30yes.

00:37:30So LTX 2.5 with accompanying audio,

00:37:34when you upload it,

00:37:36then,

00:37:36uh,

00:37:37the video has a perfect lip sync to,

00:37:41uh,

00:37:42to the audio you provided.

00:37:43it.

00:37:50Let's continue here.

00:38:07So,

00:38:08and that was for part one of that video.

00:38:20and yes,

00:38:22so that's the great thing about LTX 2.5 and Mini Max H3.

00:38:27They can do lip syncing to audio you provide.

00:38:32So that's a huge plus.

00:38:33I think other models like C-Dense 2 can that as well,

00:38:37but,

00:38:37um,

00:38:39well,

00:38:39at least it doesn't work on Kase.

00:38:40Uh,

00:38:41there,

00:38:42the image,

00:38:43the video finished.

00:38:44So you can see that he inserted,

00:38:47uh,

00:38:48well,

00:38:48he,

00:38:49inserted the spark there.

00:38:51And so that's,

00:38:52um,

00:38:54yes.

00:38:55And the robot comes to life.

00:38:57So,

00:38:58um,

00:38:58the reason why I used,

00:38:59uh,

00:39:00C-Dense 1.5 for that clip and,

00:39:03uh,

00:39:04not C-Dense 2,

00:39:05C-Dense 2 or 2.5.

00:39:07Well,

00:39:07simple.

00:39:08It's cost.

00:39:10Um,

00:39:13C-Dense 1.5 is pretty cheap on Kase,

00:39:16while the others are,

00:39:17well,

00:39:18newer models are very expensive.

00:39:21And so I'm,

00:39:22uh,

00:39:23using,

00:39:24uh,

00:39:26cheaper models for stuff that,

00:39:28well,

00:39:28isn't,

00:39:29maybe where quality or really,

00:39:32or really good movement or so is not required.

00:39:35Um,

00:39:38so if you want,

00:39:39um,

00:39:40multiple shots with a high action or so,

00:39:43then you definitely need to use one of the newer models,

00:39:46but for,

00:39:47um,

00:39:49what will be inserted and robot coming to life.

00:39:52Um,

00:39:52yeah,

00:39:53that's,

00:39:53uh,

00:39:54something pretty simple.

00:39:56So I'm using cheap model for that.

00:39:58Uh,

00:39:59now I opened my video editing tool,

00:40:01uh,

00:40:02DaVinci Resolve.

00:40:04and,

00:40:05um,

00:40:07uh,

00:40:07yes.

00:40:09Um,

00:40:13uh,

00:40:13yeah,

00:40:14I switched from VSDC

00:40:17to,

00:40:17uh,

00:40:18Resolve again.

00:40:19Um,

00:40:23uh,

00:40:24simply because,

00:40:25uh,

00:40:27VSDC,

00:40:27uh,

00:40:28costs something.

00:40:29And,

00:40:30well,

00:40:30I'm a free user,

00:40:32so I haven't paid anything yet for any AI tools.

00:40:38and,

00:40:38uh,

00:40:38when I bought VSDC,

00:40:40I think,

00:40:41okay,

00:40:41it was a one-time purchase,

00:40:42but then I found out that,

00:40:44nope,

00:40:44it was a subscription.

00:40:46Um,

00:40:46I initially bought it,

00:40:49um,

00:40:50because,

00:40:51uh,

00:40:52it could do stuff like chroma keying,

00:40:55which at that time,

00:40:56I don't know,

00:40:57DaVinci wasn't capable of,

00:40:59or,

00:40:59I didn't knew how to do that.

00:41:02Um,

00:41:03but,

00:41:04uh,

00:41:04yeah,

00:41:04DaVinci can do that.

00:41:06And,

00:41:06uh,

00:41:07also it has way more effects and everything.

00:41:09So,

00:41:10um,

00:41:11yes.

00:41:13And,

00:41:14yeah,

00:41:14I'm just

00:41:16putting,

00:41:17uh,

00:41:18the,

00:41:20um,

00:41:24the,

00:41:24uh,

00:41:24I'm renaming the image,

00:41:26the video files,

00:41:27and then I'm inserting them into the resolve.

00:41:31So,

00:41:31basically,

00:41:32I want to start with Ioni walking there.

00:41:36Um,

00:41:36I prepared,

00:41:37um,

00:41:38my resolve with the subtitles already.

00:41:41Um,

00:41:43uh,

00:41:43yes.

00:41:44So,

00:41:44as you can see,

00:41:45the subtitles also match that,

00:41:47uh,

00:41:48to the,

00:41:49uh,

00:41:50uh,

00:41:50to the,

00:41:50uh,

00:41:50to the song and the lyrics.

00:41:57Um,

00:41:59yes.

00:42:00Uh,

00:42:04I'm not really sure what I was talking about there.

00:42:09so,

00:42:10ah,

00:42:10yes,

00:42:10I've,

00:42:10yes,

00:42:11I've set up,

00:42:12um,

00:42:12the timeline already.

00:42:13So I set up the ending,

00:42:15the outro,

00:42:16um,

00:42:17uh,

00:42:20yeah,

00:42:21my whole,

00:42:22um,

00:42:23project in resolve already.

00:42:26so that I can now just.

00:42:31Let me mute that.

00:42:35Um,

00:42:35uh,

00:42:35I think at that time only Kai and me were left,

00:42:39I believe.

00:42:40Uh,

00:42:41Oh,

00:42:42we talked for 40 minutes.

00:42:43Okay.

00:42:44Um,

00:42:46and,

00:42:47uh,

00:42:47yes.

00:42:49Um,

00:42:51uh,

00:42:52I muted it now so that we,

00:42:55uh,

00:42:57uh,

00:42:57might be,

00:42:57uh,

00:42:58uh,

00:42:59because,

00:43:00uh,

00:43:00I'm not really sure when Kai will say something.

00:43:03So I don't want to speak against him in that recording.

00:43:06Um,

00:43:07yeah,

00:43:07but what you can do is,

00:43:08uh,

00:43:08simply using,

00:43:10uh,

00:43:11you can insert,

00:43:11um,

00:43:15uh,

00:43:16uh,

00:43:16using time code,

00:43:17which I just showed,

00:43:18and then they match perfectly into the,

00:43:21um,

00:43:23yeah,

00:43:23video.

00:43:24Uh,

00:43:24let's just skip a bit ahead.

00:43:26I think I was explaining a lot of stuff about DaVinci at the time.

00:43:30Um,

00:43:35uh,

00:43:35from Maestro.

00:43:36Um,

00:43:37uh,

00:43:37then yes,

00:43:39let's go back.

00:43:40And then here I was showing about,

00:43:42um,

00:43:42the models,

00:43:43uh,

00:43:43available on Casa.

00:43:44So there are a bunch of image models.

00:43:48So basically they support all image models and,

00:43:52uh,

00:43:52or current,

00:43:53uh,

00:43:53video models,

00:43:54even one 3.0,

00:43:56which just came out.

00:43:59Um,

00:44:00Freebit is the other platform I'm using,

00:44:02um,

00:44:03which has way more,

00:44:04way more video models,

00:44:06as you can see.

00:44:07So there's,

00:44:08let's say three times of them.

00:44:10Um,

00:44:12um,

00:44:12using Pixverse there a lot as well.

00:44:15Um,

00:44:17um,

00:44:17so the Pixverse is also a good model.

00:44:19Uh,

00:44:20uh,

00:44:21then I'm also using Pixverse C1,

00:44:23because the C1 model allows for short clips starting with one second,

00:44:28that no other online video tool allows.

00:44:32um,

00:44:33so one second is very,

00:44:34very quick,

00:44:35but,

00:44:36um,

00:44:36I sometimes use,

00:44:37for example,

00:44:37three seconds clip clips,

00:44:40um,

00:44:41for,

00:44:41say,

00:44:43one to,

00:44:43I have an image and I want to,

00:44:45uh,

00:44:46go from that to the next scene,

00:44:48then,

00:44:48uh,

00:44:49three second clip might be okay.

00:44:52And,

00:44:52uh,

00:44:53it's cheaper to generate a three second clip,

00:44:55then to create a five second clip and then,

00:44:57uh,

00:44:58make it play faster.

00:45:02so,

00:45:03Um,

00:45:05yeah,

00:45:05so,

00:45:06uh,

00:45:07Casa also has a new one model,

00:45:09which was,

00:45:10I think released,

00:45:11well,

00:45:11yesterday,

00:45:12so 24th.

00:45:14I haven't tried it yet,

00:45:15um,

00:45:16but,

00:45:16it's probably good,

00:45:17I believe.

00:45:19Um,

00:45:20the,

00:45:20all the other one models were quite good.

00:45:23And yeah,

00:45:24they also have the minimax models.

00:45:26Um,

00:45:28as I said in the beginning,

00:45:30minimax,

00:45:31uh,

00:45:33isn't basically allowed to be used in the EU.

00:45:36Um,

00:45:37you can get a license.

00:45:39Um,

00:45:41but I don't know.

00:45:44I hope Casa has one.

00:45:45I don't know.

00:45:46Um,

00:45:48so you can see the license here and there it says,

00:45:51um,

00:45:51uh,

00:45:54excluded territories,

00:45:56uh,

00:45:56include the European Union,

00:45:57United Kingdom,

00:45:59Korea,

00:46:00and,

00:46:00uh,

00:46:01the USA.

00:46:04Um,

00:46:05if you want to use it in these territories,

00:46:07you can contact them to get a license.

00:46:10Um,

00:46:11I don't know.

00:46:11Maybe I just do that.

00:46:12And I don't know if we have to pay something for that.

00:46:16I have,

00:46:17uh,

00:46:17no idea if they want to have money for that.

00:46:20because basically the,

00:46:22uh,

00:46:23model is,

00:46:23yeah,

00:46:23free to download and you can use it anywhere else in the world.

00:46:28Um,

00:46:31but minimax,

00:46:31minimax,

00:46:32actually is very,

00:46:33very strong model,

00:46:35especially since it's,

00:46:36yeah,

00:46:37uh,

00:46:37you can use it on your own PC and it has quality.

00:46:41That is,

00:46:42uh,

00:46:43not,

00:46:43not completely on par with,

00:46:45let's say,

00:46:45see dance too,

00:46:47but,

00:46:48um,

00:46:49it,

00:46:49it's not that far away.

00:46:51And as you,

00:46:51as you can use your own audio and it does lip syncing,

00:46:55um,

00:46:56for that.

00:46:56So that's very,

00:46:57very strong.

00:46:59Um,

00:47:01uh,

00:47:02I think I will share.

00:47:05Let's go ahead.

00:47:14I was spending a little bit,

00:47:18uh,

00:47:19then I,

00:47:19I'm showing another,

00:47:21um,

00:47:22image creation tool.

00:47:24Um,

00:47:25well,

00:47:25not only images on,

00:47:27uh,

00:47:28Google flow.

00:47:28You can also upload your characters.

00:47:31You can create videos.

00:47:32Um,

00:47:33so it's really good platform.

00:47:35Um,

00:47:36but unfortunately for everything besides images,

00:47:39you basically need a subscription.

00:47:42Although,

00:47:43um,

00:47:43you get free credits per day,

00:47:46I believe,

00:47:47um,

00:47:49to also use,

00:47:51uh,

00:47:51you can use it for videos.

00:47:53And,

00:47:53uh,

00:47:54they also have,

00:47:55uh,

00:47:56Google flow music.

00:47:57Uh,

00:47:57the former producer AI,

00:47:59which you can also use to create,

00:48:01uh,

00:48:01music.

00:48:02Yes.

00:48:03Uh,

00:48:06credits.

00:48:07Um,

00:48:08which I believe I can use also to create videos.

00:48:12Um,

00:48:14image creation,

00:48:15creating images is free.

00:48:16So using nano banana,

00:48:19uh,

00:48:19two pro,

00:48:20uh,

00:48:21two light.

00:48:22I believe the nano banana one is also available.

00:48:26Um,

00:48:30so,

00:48:31um,

00:48:33uh,

00:48:34yes.

00:48:35So,

00:48:35yeah,

00:48:35what you can do,

00:48:36um,

00:48:37with Google is for example,

00:48:39um,

00:48:40yeah,

00:48:41um,

00:48:41you can apply,

00:48:42upload a screenshot.

00:48:44Then I told it to add,

00:48:47uh,

00:48:47text for YouTube thumbnail.

00:48:49And then I told it to change the aspect ratio to nine to 16,

00:48:53uh,

00:48:54for TikTok thumbnail.

00:48:56And it did that.

00:48:57So,

00:48:58and completely free.

00:49:02Um,

00:49:03the comments or somewhere.

00:49:06Um,

00:49:07um,

00:49:07Ah,

00:49:07then,

00:49:08um,

00:49:09we,

00:49:10uh,

00:49:11I went back to meta and found out that,

00:49:13um,

00:49:14the,

00:49:15uh,

00:49:16images created,

00:49:18uh,

00:49:19successfully.

00:49:19Um,

00:49:20I think I can skip that part ahead.

00:49:24Um,

00:49:24we were a little bit lost about the sorting,

00:49:28uh,

00:49:28of the images.

00:49:29So here's seven and here's 31.

00:49:32So we thought,

00:49:32Hey,

00:49:32what's going on?

00:49:34but once we figured out the,

00:49:35the sorting,

00:49:37and then it was,

00:49:38uh,

00:49:38really good.

00:49:40Um,

00:49:41so I'm skipping ahead a bit.

00:49:43Um,

00:49:44then,

00:49:46uh,

00:49:47uh,

00:49:47what I wanted to share next was,

00:49:50um,

00:49:51that you might need to have different tools for different stuff.

00:49:57For example,

00:49:59for the Sanctum Ovalis music video,

00:50:02I wanted,

00:50:03um,

00:50:04the cultists to,

00:50:05to,

00:50:06uh,

00:50:06open their surprise eggs and,

00:50:09um,

00:50:10present chocolate and also,

00:50:13um,

00:50:15the,

00:50:16uh,

00:50:17aluminum foil where the surprise eggs were into,

00:50:20uh,

00:50:21should match the color of the original eggs.

00:50:26Um,

00:50:28that was something that was quite difficult to do.

00:50:31So,

00:50:32yeah.

00:50:32So you can see there the,

00:50:34then I'm,

00:50:34I'm opening the original image.

00:50:39Um,

00:50:42um,

00:50:44but,

00:50:44uh,

00:50:45yes,

00:50:46that one,

00:50:46for example.

00:50:47So you can see,

00:50:48for example,

00:50:49the second from the right,

00:50:50the blue one,

00:50:51the blue egg,

00:50:51um,

00:50:53should have was,

00:50:54of course,

00:50:54blue aluminum foil,

00:50:56uh,

00:50:57there.

00:50:58Um,

00:51:00and yeah,

00:51:01I went to Meta and told it to do that.

00:51:05Um,

00:51:08so,

00:51:08and as you can see there,

00:51:10the aluminum foil is purple.

00:51:12Um,

00:51:13yeah,

00:51:14then,

00:51:16yeah,

00:51:16I told it to do that,

00:51:18but,

00:51:19uh,

00:51:20it didn't really match on every,

00:51:22uh,

00:51:23account.

00:51:24So that's why I then went to,

00:51:26uh,

00:51:26Google,

00:51:27uh,

00:51:28and,

00:51:29uh,

00:51:29did that,

00:51:30uh,

00:51:30there,

00:51:30because Google flow was better at,

00:51:33well,

00:51:34it's probably better at,

00:51:36uh,

00:51:36creating,

00:51:37matching aluminum foil.

00:51:40Um,

00:51:43yes,

00:51:44um,

00:51:46but Meta,

00:51:46for example,

00:51:47is better at creating,

00:51:48German text,

00:51:49because Google has,

00:51:52um,

00:51:52likes to translate the German text into English,

00:51:55and then back to German,

00:51:57and that it's not always the same text you wrote.

00:52:00Um,

00:52:01yes,

00:52:02or

00:52:03it

00:52:04writes English,

00:52:05the English translation,

00:52:06even if you said,

00:52:07please write the text in German.

00:52:09So you always get the right tool.

00:52:12Um,

00:52:12you might have,

00:52:13um,

00:52:14yeah,

00:52:14to,

00:52:16try to give it a few tries.

00:52:19until you find the right tool.

00:52:21So,

00:52:22um,

00:52:22here as well.

00:52:23Um,

00:52:24that's,

00:52:24um,

00:52:24where I'm going back to the foil.

00:52:26So you can see,

00:52:27um,

00:52:29there are multiple capsules as well,

00:52:32in these,

00:52:33um,

00:52:34yeah,

00:52:35in the,

00:52:36uh,

00:52:36chocolate X,

00:52:37but,

00:52:37in surprise X,

00:52:38there's only one capsule.

00:52:39so I told it to,

00:52:42I only want one capsule.

00:52:45And then I,

00:52:46yeah,

00:52:47with the help of,

00:52:49Lando Banana 2,

00:52:50um,

00:52:51I also managed to get the aluminum foil into the correct color.

00:52:55Then,

00:52:56um,

00:52:56well,

00:52:57there's saying there's a figurine in the,

00:52:58every seventh egg.

00:53:00If you see now that was in the sixth egg.

00:53:03So,

00:53:03I told it,

00:53:04I want the figurine in the seventh.

00:53:06Um,

00:53:06then it,

00:53:07I don't know what did to put a Gundam or something like that into the sixth egg.

00:53:12Um,

00:53:13yes.

00:53:14So,

00:53:14um,

00:53:16also things like these.

00:53:18Uh,

00:53:18yeah.

00:53:19And then I,

00:53:20uh,

00:53:20it took a few attempts.

00:53:22That was one of them.

00:53:24I don't know what it did there.

00:53:27Um,

00:53:28took a few attempts.

00:53:28So,

00:53:29no,

00:53:30there's

00:53:31an overflowing egg.

00:53:34And,

00:53:35um,

00:53:36yeah,

00:53:37I think I created,

00:53:39uh,

00:53:39the last one then again with meta.

00:53:41To get,

00:53:42um,

00:53:44the,

00:53:44well,

00:53:45egg,

00:53:45the,

00:53:46uh,

00:53:47contents of the egg,

00:53:48uh,

00:53:50appropriate.

00:53:51So that was the figurine.

00:53:52Yes.

00:53:53Yes.

00:53:53Mini figurine.

00:53:54And there were some dinosaurs as well.

00:53:57Um,

00:53:58yes.

00:53:59But yeah,

00:54:00that's only with a colorful,

00:54:03uh,

00:54:05drain.

00:54:12So yeah,

00:54:13you might have to try it out on different,

00:54:15uh,

00:54:16tools.

00:54:16If you can't get it to work,

00:54:18um,

00:54:19after a few attempts,

00:54:20attempts on,

00:54:21on,

00:54:22uh,

00:54:22your first or two,

00:54:24even sometimes in your chat,

00:54:26might also help.

00:54:27So to basically to reset everything.

00:54:30Uh,

00:54:31let's see.

00:54:32I don't even know what I was doing now.

00:54:35Um,

00:54:37oop.

00:54:45So I think we were talking about the,

00:54:49uh,

00:54:50models.

00:54:51Um,

00:54:52and yeah,

00:54:53the,

00:54:53I think we'd talking a bit over,

00:54:55about Kai's hardware,

00:54:57how much,

00:54:58uh,

00:55:00uh,

00:55:00what,

00:55:01what his hardware might be capable to be doing.

00:55:03Um,

00:55:08yes.

00:55:09Um,

00:55:11then I think I opened the,

00:55:13ah,

00:55:13right.

00:55:14Uh,

00:55:14then we checked it.

00:55:15My scroll right now uses around 200 gigabyte,

00:55:18uh,

00:55:19with all the models I have.

00:55:20So,

00:55:20you need to have a lot of disk space if you want to do,

00:55:25want to create AI stuff locally.

00:55:28Um,

00:55:31um,

00:55:31for Pinocchio right now.

00:55:34A little bit below.

00:55:35Yes.

00:55:36Yeah.

00:55:38Um,

00:55:43I think I was talking at the same time.

00:55:45So talking at clicking doesn't seem to be by strong,

00:55:48uh,

00:55:48point.

00:55:49And yeah,

00:55:50here are the checkpoints.

00:55:51as you can see 165 gigabyte in,

00:55:55uh,

00:55:56video models.

00:55:57And the rest is,

00:55:57uh,

00:55:58let's say created videos and stuff.

00:56:00And,

00:56:00uh,

00:56:01yeah,

00:56:01there we can see,

00:56:02I start with ATX 2.3 in my scroll.

00:56:06Um,

00:56:07the,

00:56:08uh,

00:56:08Minimax models are 20 gigabytes.

00:56:10Um,

00:56:11I've got a special,

00:56:13uh,

00:56:13version of the ATX 2.5 model,

00:56:17uh,

00:56:17which is pretty small.

00:56:18So 14.3 gigabyte,

00:56:20uh,

00:56:21that is optimized for RTX 50 cards.

00:56:24So that's smaller than the regular model by,

00:56:27and using the same quality,

00:56:30um,

00:56:31or,

00:56:31uh,

00:56:31delivering the same quality.

00:56:32So quite a good deal.

00:56:34If you have a RTX 50 card,

00:56:36like me,

00:56:37um,

00:56:38which I luckily got last year for 700 euros.

00:56:41Now it costs twice as much.

00:56:44Lucky me,

00:56:45sorry guys,

00:56:46I won't sell it.

00:56:47Um,

00:56:51let's go ahead.

00:56:53So yeah.

00:56:53Um,

00:56:54the Minimax models are around 20 gigabyte.

00:56:57I have a quen image edit model,

00:56:59um,

00:57:00to create images,

00:57:01which is 20 gigabyte.

00:57:02Um,

00:57:05um,

00:57:06um,

00:57:07yeah,

00:57:08then

00:57:08in the model browser,

00:57:10you can basically download all models,

00:57:14that exists,

00:57:16the one models,

00:57:17um,

00:57:18lots of stuff.

00:57:19So Maestro is a really,

00:57:20really powerful tool.

00:57:22Uh,

00:57:23let me just quickly skip ahead.

00:57:26So,

00:57:27ah,

00:57:27and behind Maestro is also,

00:57:30if you open the classic UI,

00:57:32the,

00:57:33uh,

00:57:33well-known one,

00:57:34two degree UI.

00:57:36Um,

00:57:37so,

00:57:38uh,

00:57:38Maestro is built upon one,

00:57:40two degree.

00:57:41Um,

00:57:42I've used that a lot and probably other,

00:57:44you are,

00:57:45might,

00:57:46might have seen that as well.

00:57:47So basically that one,

00:57:49two degree was the,

00:57:50the tool.

00:57:51If you wanted to go local video generation,

00:57:54aside from Confu UI,

00:57:57I hate Confu UI cause,

00:57:59if you get something to work in there is a nightmare.

00:58:03Um,

00:58:03every custom node you need is incompatible to be at least 10 others.

00:58:07So,

00:58:07yeah,

00:58:08that's why I love the,

00:58:10uh,

00:58:11one two two P interface because you don't need to haggle around with that.

00:58:16Um,

00:58:17and I love even Pinocchio even more using that because you don't need to,

00:58:21have to try to get that working as well.

00:58:24So just,

00:58:25it's a one click installer,

00:58:27install one two two P or install Maestro for me and then you're done.

00:58:32So,

00:58:34um,

00:58:35uh,

00:58:36then I was quickly showing some images I created for,

00:58:40uh,

00:58:41for the band fest,

00:58:42um,

00:58:43the band camp festival.

00:58:44And these are created using,

00:58:46uh,

00:58:47the,

00:58:47uh,

00:58:49agent with,

00:58:50uh,

00:58:50within Maestro.

00:58:51and I don't know,

00:58:53basically the imaging model should be strong.

00:58:55I,

00:58:56but I'm not sure what the agent does,

00:58:58uh,

00:58:58So that doesn't look that good.

00:59:02Um,

00:59:03so,

00:59:04yeah,

00:59:05I've tried the agent within Maestro,

00:59:08but,

00:59:09um,

00:59:09I mean,

00:59:10these are the images it creates and out of these images,

00:59:14then also creates the videos.

00:59:16So,

00:59:18yeah,

00:59:18if the image is bad,

00:59:19then the video will also be bad.

00:59:21So,

00:59:22um,

00:59:23yeah,

00:59:23I'm not really,

00:59:24uh,

00:59:25using the agent anymore.

00:59:27Um,

00:59:29and,

00:59:30yes,

00:59:31I believe then,

00:59:33uh,

00:59:33we are also at,

00:59:34uh,

00:59:35the end.

00:59:36So then I figured out,

00:59:38okay,

00:59:38uh,

00:59:3820 minutes were missing from my video as well.

00:59:41Um,

00:59:42uh,

00:59:43yeah,

00:59:43I,

00:59:44uh,

00:59:45made a short,

00:59:46uh,

00:59:48section with that.

00:59:49So,

00:59:50um,

00:59:51yes,

00:59:51um,

00:59:52then I'd say,

00:59:54thanks for listening.

00:59:55Thanks for,

00:59:55thanks for watching.

00:59:57Um,

00:59:57and,

00:59:58um,

00:59:59I've been asked to for another session.

01:00:01Uh,

01:00:01yes,

01:00:02we can do that.

01:00:02Um,

01:00:04yeah,

01:00:04take care.

01:00:04Have a nice day.

01:00:05Bye.

Making an AI Music Video from Scratch Full Tutorial & BTS Workflow)

Kategorie

Transkript

Ione erklärt die Welt von #Uniworld

Empfohlen