I don't think there's any need to upscale the images in Gigapixel AI if you've already upscaled them x4 in Midjourney. 4x Midjourney images are 5824 x 3264 px. All you would need to do then is size down the image to 3840 x 2160 px and adjust the PPI in Photoshop or any other photo editing software. The quality of the upscaled images in Midjourney is quite good in my opinion. I only use Gigapixel AI when I have a high number of images to upscale just because Gigapixel allows for batch processing and I find it quicker than Midjourney but I would say that the quality of upscales in Midjourney is actually superior to Gigapixel AI in most occasions.

​@@EyEyeArt makes sense. what I do now is I create an image in midjourney v6, when I pick the image I like I use vary regions to fix all small details and also change midjourney to v 5.2, so I get upscale x4 available, After I upscale it in midjourney x4, and I adjust the image in Photoshop to 16:9 ar, 300 dpi (3840 to 2160 resolution). As a final touch I use gigapixel to smooth the image after Photoshop adjustments. I only worry about should I sell png files or jpeg? I see a lot of people sell the arts in jpeg, but png quality is better...

I don't use Photopea very often but I would say you could probably upscale an image by using the 'Image Size' tool from the 'Image' tab in the main top menu bar just like you would in Photoshop, although, I wouldn't trust the results. Even in Photoshop this doesn't normally produce very good outcomes.

No, absolutely. It's just that JPG files are smaller, more manageable and easier to share.

Thank you very much! I'm glad you appreciate the content!

Thank you for sharing your thoughts. I will try my best to get p.2 out ASAP... also working on a bonus p.3 ;)

Hey, I understand where you're coming from and I agree to an extent, but have a look at this video if you think that scanning AI songs for copyright infringement is overkill: kzbin.info/www/bejne/fXyslX5_gb6ekLM

Thank you very much! Glad you enjoyed the content.

Thank you very much! I'm glad you found the content useful.

Thanks a million! I’m glad you appreciate the content. As regards your queries. I’ve never printed interior wallpaper so I’m not sure what the requirements for that would be. I would liaise with the printing facility to see what their requirements for printing wallpaper are. There are lots of POD companies that do wallpaper nowadays. I would reach out to them or search their websites to get an idea of the ideal resolution your images should be. I did find this resource online: www.contrado.com/bespoke-wallpaper. If you scroll down to ‘Help’ and then ‘Print Guideline’ you’ll find some information on the image resolution and file formats they suggest. You can also click on the ‘Start Design’ button, choose ‘Photo or Collage Wallpaper’, input your required wallpaper dimensions and upload your image to see if their system’s detectors consider your file of good quality. This website also gives some information on the image file type and resolution they require under the ‘Design File Info’ tab: printedmint.com/shop/wallpaper-242#attr=428. Although, this seems to be for repeated pattern wallpaper rather than a photo wallpaper that I think is what you’re after. I would still contact the printing facilities just to be sure, though. If they’re serious about their clients and the quality of their products they should provide the information you need. Gigapixel AI is, in my opinion, industry standard when it comes to upscaling images although it does require a lot of resources from your device because it runs locally on your computer, especially if you intend to carry out ‘extreme upscaling’ as you defined it, get ready for very long processing times depending on the computing power of your PC or laptop. I know you can rent out GPU units online to use locally on your computer to enhance your device’s performance but I’ve never done that myself so I wouldn’t be able to give any suggestions on that.

Yeah, I think you're right. As I mentioned in the video, the software is still in its beta version and doesn't seem to be very capable of detecting potential infringement at the moment but I thought it was worth mentioning since, in my opinion, it seems like it's got great potential.

@@EyEyeArt Definitely, and I applaud the initiative. I just hope people don't rely on it too much.

Thank you very much! I’m glad you appreciate the content. I’ve been running into the same issue lately and after finding nothing online to help understand why this is happening I think I figured it out. The problem is most likely the way the software is programmed to round off the dimensions of the image to the nearest number. If your multiplying factor is a decimal number (e.g. 2.647) and not a whole number the software will have to decide to round both height and width either up or down to nearest whole number, but it will only go one way for both values: 1451 x 2.647 = 3,840.797 px 816 X 2.647 = 2,159.952 px In this case I noticed the software will round both values down to 3840 x 2159 px. A workaround would be to size down the image (while still maintaining the aspect ratio), before upscaling in Gigapixel AI, to a value that can be multiplied by an upscaling factor that is a whole number. You can do that by dividing your desired output dimensions (e.g. 3840 px and 2160 px) by 2 or 3. In this case you would start out with images that are either 1920 x 1080 pixels and upscaling them x2 or 1280 x 720 pixels and upscaling them x3. 1920 x 2 = 3840 px 1080 x 2 = 2160 px 1280 x 3 = 3840 px 720 x 3 = 2160 px If you want to start off with the higher resolution (1920 x 1080) then you could consider upscaling x2 in Midjourney before downloading and resizing the image files. I noticed that this ‘glitch’ will also occur with other image processing software (like Photoshop) so you would probably be better off resizing the files early on in your workflows, maybe straight after downloading them from Midjourney, if that’s what you’re using to generate your images.

I touched on this in the video as well. If you trust these music generators' built-in moderation (although the moderation errors normally just refer to citing an artist's name in the prompt from what I understand), then, of course, this is redundant. It was just a way for me to have some peace of mind since I wanted to publish this video on KZbin. By no means, though, would these methods be definitive in detecting plagiarism, anyway. As to getting the song heard, I would say it's equally hard to get any newly composed song heard whether it's AI generated or not.

Thank you very much! I'm glad you appreciated the video.

Thank you very much! I'm glad you found some useful information in the video. I'm hoping to get part 2 out soon.

Thank you very much! I'm glad you enjoyed the video. Part 2 is scripted already. I just have to find the time to screencast, edit and publish it. Hopefully soon. 1. The artifacts, to me are sort of plastic, artificial looking parts of the image, that don't resemble actual brush strokes or a paper/canvas surface. You can see them when you zoom in on the image. It looks like the image was painted on plastic wrap rather than regular paper/canvas. If you're printing the images, the physical surface of the paper or canvas would most likely make them less noticeable, whereas on widescreen TVs, that can be seen from a close distance, these artifacts (again anything that doesn't closely resemble an actual brushstroke or paper/canvas surface) may be more noticeable. 2. Of all the trials I've carried out so far, in my opinion, Midjourney works best when trying to reproduce vintage paintings, although, AI image generators are advancing at an extremely fast pace so I would image that, if other AI image generating platforms aren't already capable of getting similar results, they will most certainly be able to very soon. Also note that certain AI models specialize (or excel) in generating only certain types of images so this may also be something you may want to consider when looking for an alternative to Midjourney.

@NewMexicoCountrySongs I'm glad you enjoyed the video. Thank you very much! I appreciate it greatly!

I agree completely. Artistic expression has existed for centuries, millennia even, probably from the dawning of human civilization without the necessity of attributing proprietary rights to any of it. The music industry and copyright laws are modern constructs that, in the recent past, have proven to be unable to constrain any kind of artistic expression to proprietary laws and corporate interests. I remember the times when it was so hard to publish independent music, or discover independent music, since all the major record labels owned, and dominated the music broadcasting and marketing outlets. I would have to program the recording of music cassettes on my Pioneer stereo to manage to listen to those few nighttime radio stations that would broadcast alternative/indie bands. That obviously all changed recently with the advent of KZbin and current streaming services and corporate music had no choice but to loosen their stronghold since they understood there was no way they could stop people from share music online anyways. I'm not saying, let's get rid of a system that justly retributes artists for original songwriting. I understand that musicians livelihoods are at stake here. Nevertheless, considering the incredible possibilities for artistic discovery and experimentation that AI technology offers all of us, I don't see it succumbing to a group of artists or producers that are more interested in its financial repercussions rather than its artistic outcomes. intellectual Property laws are proven to be fundamentally flawed and will necessarily have to adapt to the everchanging music and arts scenes as they progress and evolve.

​​@@EyEyeArt I think you will find that music generated with the use of AI will not be played on radio, tv or Spotify in the very near future. Since it is not copyright capable, it will be patched into by broadcast and streaming platforms at first, because they will see that they don't have to pay performing rights organizations, saving massive amounts of money. What happens next is massive boycott of these streaming platforms and broadcasters by all the major recording artists. That will create an opening for broadcasters/streamers who vow to not allow any AI generative music on their platform. So if you want to hear the new Taylor Swift album or any other artist, you will have to go to that non-AI platform. Other platforms will fall in line and either only have more and more variations of the same music used for trained AI or go to human based music. But.... I am amazed at what AI can do to simulate human qualities in music.

People that are experimenting with AI generated music and artwork don't seem to be all that interested in copyright anyway. If you really want to, you can still generally monetize the AI works without them being protected by intellectual property. The purpose of this video, wasn't to check for plagiarism because I wanted to copyright the tune myself. I just wanted to be sure I would be able to safely make it public here on KZbin, or anywhere else on the internet, without being struck by a copyright claim. I would like to clarify that when I refer to plagiarism it's in the traditional sense of publishing music that is a direct citation or copy of a previously published composition and passing it for one's own. There is no legislature, at the moment, to support the idea that training AI models with copyrighted music constitutes plagiarism. It's just speculation.

Is that illegal to train your software that way? 🤔

@Musicmanbutte no, I let Udio do everything for me

@@EyEyeArt I write all of my own lyrics. Here is one I just created tonight. It was written about my other that passed away Dec -19, 2017, she passed away 14 days after her 76th birthday. Here is the link if you want to listen to it. I hope you don't mind me posting a link to this song. I'm just posting it for you to hear. I am a paid member of Suno. kzbin.info/www/bejne/fmbLo2ylesZroqM

@Musicmanbutte Thanks for sharing

@Musicmanbutte thanks for pointing that out, although I do greet everyone at the beginning of all my videos, including this one, and welcome them (back) to the channel. You may have missed it probably because I like to keep the greetings very brief and get straight into the content.

@Toxicflu I think it's just a matter of time. It's probably the reason why Suno's voices sound so processed and artificial. They may still be trying to figure it out. Singers tend to all have a very similar way of singing within the same genre. Look at Grunge singers during the 90's or Heavy Metal singers. I think AI developers will eventually find a way to emulate a certain style of singing without sounding too similar to a specific artist. Udio is still in beta testing at the moment. If we look at the progress generative AI has made in image generation, in such short amount of time, I think we can still expect pretty surprising results from this kind of technology. Just for the record, I'm not happy about AI generated voices sounding like known singers or even the possibility of using a singer's voice without consent. Although, regarding the idea of a singer's voice likeness being copyrighted, if laws like that were ever to be implemented then bands like Greta Van Fleet couldn't exist and I'm not sure that would be fair either. I should have mentioned this in the video but I'm certain that our ears are probably one of the best plagiarism detectors we have and if a any part of a song sounds too familiar or too similar to some other song, then I wouldn't pursue it. In my AI song, I thought the general vibe during the piano solo sounded a lot like Norah Jones. But then again even traditional musicians aren't creating anything from scratch so, I think that any newly composed song has to have some level of familiarity and similarity to existing music. Actually that's probably part of its allure.

There was a guy that made a software to create every midi note combination and sent all of them to a copyright office to claim every note combination. And they replied that the songs were not significant enough to be copyrightable. Thank god!

@@Toxicflu I hope that's what I'm thinking of and misremembering. Was it a few years ago? If so, that's probably it. And like you said, thank god. Nobody needs that sh!t. LOL.

@ParvathyKapoor like I said in the video, the AI has access to enormous datasets of original music during its training so I would guess there should be a way to align the model to avoid copying any music it was trained on ...at least you would think.

@wedontexist369 Udio is still in beta testing. I'm sure they'll figure it out. They seem to moving fast with all the new features they've added recently. I'm positive the next version of their AI model will be a big upgrade from the current one and they will have probably already addressed these issues by then.

There are over 67 billion combinations of 8 notes at the same length within one octave. I think we have a while to run out of combinations.

@@Toxicflu yeah I think he made more than a billion hr mad enough so any song u can think of is his and he copyrighted it all

​@@ToxicfluLuckily its enough to write those melodies down to actually be the copyright holder. And that is what his algorithm had done for all of the melodies of the natural notes. 😅

No one should be able to copyright something everyone can do . making a melody , what happens when melodic combinations run out and no1 can use any melody? F*** copyright.

Thats F****** Stupid

So as every art is. Only God creates ex nihilo.

Consider that this is the beta version of the website. I'm sure that as they refine their training and new versions of their AI model are rolled out, these issues will be dealt with as well. I’m not sure if you’re familiar with AI image generation but if you compare the very first versions of the Midjourney AI model to the most recent iterations, you can have an idea of the level of progress that can be made in such a short period of time.

@Esteban-pb4gw you are correct, I agree. I have no intention of registering any AI generated song for copyright. I just wanted to check if any part of the song that was generated by Udio may have been copied from an existing, published song to avoid any possible copyright infringement claims since I'm making the song public on KZbin.

@@EyEyeArt well to me, at some point it is a copyright infringement from the minute Udio and others train their model from your songs, your voice, your style of writing. It's a big mess in this department really

@Esteban-pb4gw at one point I'm sure it will become clearer to all of us. I guess we'll just have to ride it out and see what happens.

The part that you edited, or changed can be though. But how much is needed is still in debate. aka, is just adding a shaker, or a reverb enough???

@Alants8 the purpose of the video is to check if any part of the song that was generated by Udio may have been copied from an existing, published song. It's just a record of myself exploring some current plagiarism detection technologies to avoid incurring copyright claims by making the song public on KZbin. I agree, and it’s understandable, that just the fact that two songs have the same or similar chord progressions doesn’t mean there’s a case for plagiarism although, in the past, lawsuits have been filed on those grounds. See the most recent Ed Sheeran’s vs. Marvin Gaye trial. I’m not sure what you mean by wanting to copyright the song or a particular chord progression but since this was generated by Udio I don’t own the rights to the song, I just want to be sure no one can come after me for using it without permission, in the case, any part of the song was mistakenly copied by the AI generator. For me this is part of a journey to discover what’s possible using AI technology. I’m really just having fun trying to see what I can come up with, although I'm also being careful not to infringe anyone’s intellectual property in the process.

@Alants8 No one is arguing that, the problem is training a model on copyrighted works without consent or compensation. It’s that simple. Good luck trying to copyright an udio output.

​@@BenCaesar Udio, like a camera is a tool. And photographers can copyright their pictures, even though they only pressed one button. Prompt composing is actually more than that. As of now, the part of the ai song that you edited afterwards, or changed can be copyrighted. But how much change needed is still in debate. Is just adding as little as a shaker, or a reverb enough???

yeah. I don't own C major

There is way more to songs than mere chord progressions. You can’t be serious.

Hi, yeah, as I was saying in the video, MIPP is still in beta testing and I don't think the results it gives can be trusted completely, at least not yet. I do think, though, that it's worth keeping an eye on since it's the only app I found that actually gives such a detailed report on which songs and namely which parts of the songs are more likely to be subject to copyright infringement.

@@EyEyeArt Okay, so I took the time to look into it and take the pdf apart. We've got a couple of problems with this one. 1. The rules for what does and does not constitute plagiarism are absurd. If you have a 4/4/8 chorus, no matter what you do, you'll always get it flagged for plagiarism. Anything rock, country, blues or pop. Also, forget about power chords. 2. Timing, bpm, genre, lyrical content, and context are not taken into account, at all. And you could sidestep this simply by changing the key of your song. 3. It converts your song into midi, (badly, you can tell by the way the pdf renders. It'll show you sections of it) or tires to. And that's how it does the analysis. But, ai music like ai images always has noise in it, which breaks the midi parsing mechanism. The effect is that you end up getting notes that are out of key, and out of time, and that's part of why your ai music scores twice to three times as high as organic music, and causes either false positives or negatives 100% of the time. A tool like this is going to have a VERY limited useful use case. My hope is that this comment is helpful and contributes what I know to the conversation.

@paulahawthorne7807 unfortunately I almost exclusively use Photoshop to carry out most of my editing. There are some open-source/free alternatives to Photoshop though like Gimp - www.gimp.org/ for desktop and Photopea - www.photopea.com/ that is an online editor with a very similar interface to Photoshop which also allows for recording actions and batch processing. To be honest, though, considering that Adobe is investing more and more resources in developing and integrating cutting edge AI features in their suite of applications, I would personally invest some time, if possible, in getting acquainted with the software. If you're looking for an alternative to Gigapixel AI you can try Upscayl - www.upscayl.org/ that is a desktop application although it looks like they're working on an online version as well, and you can also have a look at the website DGB.LOL - dgb.lol/. I can't guarantee but I'll will look into it and see if I can create a tutorial using alternatives to Photoshop and Gigapixel AI.

@@EyEyeArt Thank you so much, I am familiar with Photopea. I like it a lot.do you think this would work in Photopea?

@paulahawthorne7807 You can do all the basic image editing and I know you can create actions and batch process images in Photopea although the method is slightly different compared to Photoshop. I’m not sure about the upscaling though. It all comes down to the quality of the upscaled images. You would have to test whichever upscaler you decide to use and maybe send the upscaled images off to print to proof them. If upscaling 2x or 4x is enough you can always use Midjourney’s upscaler. You can get a 3840 x 4800 pixel image (4:5 ar) that can print up to 12 x 15 inches at 300 DPI.

Obrigado! Fico contente por apreciares o conteúdo.