Terrors From Beyond
HTML-код
- Опубликовано: 28 фев 2024
- Welcome to Xinferis TV, your gateway to television from another reality.
🔮 If you enjoyed this video, don't forget to give it a thumbs up, share your thoughts in the comments, and hit that subscribe button to join our community.
👉 Know someone who would appreciate my music? Feel free to share this video with them.
🎶 Listen to my music on:
🎧 Spotify: open.spotify.com/artist/1hQWO...
🍎 Apple Music: / xinferis
🔗 Explore more of my content and stay up-to-date with my latest projects via my Linktree: linktr.ee/xinferis
🎹 Curious about my creative process? I create the music using the following software tools:
Logic Pro
Native Instruments Komplete 14 CE
Arturia V Collection X
Arturia FX Collection 4
Arturia Pigments 4
u-he Repro-5
u-he Repro 1
u-he Diva
u-he Zebra Legacy
Ozone 11 Advanced
IK Multimedia Total Studio 3.5 MAX
🎛️ And the audio hardware below:
Native Instruments Kontrol S61 MK3
Akai MPK-49
Arturia KeyStep
Arturia BeatStep Pro
Behringer DeepMind 12
Behringer Model D
Behringer Neutron
Arturia MiniBrute
Korg Minilogue
MOTU 828es
Yamaha HS8 Studio Monitors
Yamaha HS8S Studio Subwoofer
🎨 I use the following creative tools to make the art:
Midjourney
Draw Things (which uses Stable Diffusion)
Pixelmator Pro
Apple Photos
🎬 And these editing tools to create the videos:
Apple Final Cut Pro
Apple Compressor
Apple Motion
Thank you for visiting the worlds of Xinferis TV! Your continued support helps drive me to continue creating more videos. 🚀🌌
🚫 Copyright Notice: All of the music and art on this channel are created by me. The content of this video is not royalty-free, and I reserve all rights to the video, music, and art. Any unauthorized use or reproduction of my content is strictly prohibited.
#synth #soundtrack #aiart
🖤🖤🖤
Thank you for watching!
Excellent as usual. Kudos.
Thank you!
3:25 Wooaah! flashbacks to my birth there. "Leaving...hrr...comfortable place....NUUUOOOO........Whhaaaaaaa....UuuWhhaaaa". 6:35 The actual look on my little face at said time. I always look forward to the next iteration of the 'Eve discovering the apple' goddesses you create in these videos.
Thanks for making these videos----it is always cool to see more of them!
Thank you for watching them!
당신이 창조한 객체 하나하나가 모두 아름답습니다.
모든 장면은 실제로 존재합니다.👍👍👍👍👍👍
Thank you! I really appreciate the kind words.
Man could imagine this in a movie.
I keep telling him that, but we have to give him time to work with his ideas and the tech. I think the tech also need some evolving to do, so we gotta be patient on that front, too.
Who knows. Maybe someday a producer will come with bags of money, and I could create a movie with similar aesthetics. ;)
Suberb song and the visuals are as lovely as ever, an almost Hammer Horror meets Tobe Hooper vibe, very nice! :)
Thank you!
Legend!
7:43 reminded me of that Megadeth tune "Wake up dead"
Это вы изобразили королев которые не могут решить вопрос
It's been a while since I listened to that album. I will need to have a listen again. :)
Cool videos.
I keep expecting Dr. Who (Tom Barker version) to show up.
Hah. You never know! ;)
As always, top notch, thanks!
Thank you for watching! :)
This makes me nostalgic for a 1960s British scifi movie that never existed.
:) Thank you!
aaannnndddd we're back in the realm of nightmares and disturbance - magnificent as always my friend and thank you so much for these. I wouldn't dream of asking for details of the prompts you use for these but do you make significant changes/nuances for each session or are all images the results of a singular, well thought out prompt?
It depends on the video and the image, but I typically have a base style prompt that gives the overall style. I spend a lot of time developing the base style prompts. Then I typically have a style modifier prompt I put after that. That will kind of tune the style a bit to add some variation but still have the overall same feel. Depending on the model and image I may just use those two elements and see what I get. This part of the prompt creation process is the most important because it's effectively the art style of all the images you are generating.
Often times I will also describe the scene and location. For example when generating the thumbnail for this video, I probably used something like "medium shot of a woman", although I was likely a bit more descriptive than that. Location is just where the scene is taking place. If it's outdoors I may add the time of day, like "at night" at the end.
I build up a library of these different elements so I can build different images using them in different combinations. For example "medium shot of a woman", I would know what kind of results I can get from it, so I can reuse that element with different styles and locations. I think experimenting and building a library is important, because in many cases the model doesn't create exactly what you type. You can then use those elements from your library directly or modify them to get the results you want. A good example of this is if you type "an alien planet" or even "an alien planet in space" it will give you terrain images, or possibly a planet viewed from the surface of a planet. You have to type something like "a alien planet viewed from a space ship" or something similar to that, to actually get an image of an alien planet.
An example of the general prompt format would be as follows:
[Style Prompt] [Style Modifier] [Scene Description] [Location]
With Midjourney 5.2, it's possible to get good results without being too specific. For example you can just use the Style Prompt and Style Modifier and get some interesting results. With Midjourney 6 you have to be pretty specific about what you want, or else you get some pretty extremely boring images. So with Midjourney 6, I suggest using all four elements every time. Even when developing the style prompts for Midjourney 6, the scene description and location prompts should be added to the prompt, otherwise it will be difficult to know what the results of the style prompt will be in practice.
For style prompts, one thing that works well for me is adding "using concepts from [artist, author, etc]". It doesn't replicate the art style from the artist, but it will tune the style of objects in the scene to be what the AI model thinks the artist would imagine. You can specify multiple people too. Keep in mind, every model's interpretation of an artist is completely different. So with each model you will get completely different results. Honestly a good way to think about it, is that each AI model has its own style with specific prompts.
I don't usually share exact prompts I use because specific prompts are highly tied to specific versions of models. What I stated here works on every model I have tried it on.
@@XinferisTV Amazing amount of work you put in and it's all very appreciated!! Are we close to being able to repeat generate images of a certain creation - say a monster but at different angles? Because you really need to be able to create a narrative story linking the images
It isn't possible to create multiple images of the same exact monster. It's also impossible to create different angles of the same location. The same thing with creating the same person. I have done some tricks where I could get some images that looked like the same person, but the clothing was different. Also it was only maybe 1 in 10 images generated where the person looked close enough to being the same person.
In the videos I do, I avoid as many of the things AI is bad at generating as possible. That creates a lot of restrictions. One is the on mentioned, I can't create two images of the exact same subject or place. Although I do kind of work around that and you have probably seen that in some of my videos. If it's outdoors at night for example, it could look like the same place. Or if things are taking place in a cavern, it feels like the same place overall, even though each image is effectively a different cavern.
Also AI does a pretty poor job with complex images. What I mean by that, is it does better when there is a single subject in the image. So showing one person works very well, but it typically struggles when you try to have more than one person in a scene. Sometimes having two subjects can work, but more than two typically will have a lot of issues. For example the images I have of a woman with a monster. Those actually work exceptionally well in general. The main issue with them is getting both the woman and the monster in focus.
There are some situations where I can create more than two subjects and have it work well in many cases though. One is just having the AI create a group of identical subjects. Like a group of men in cloaks. The other is one subject with different subjects behind that subject. Like a man with a group of robots behind him.
Also even though people say AI can do hands now, it does a super poor job with it. I try to avoid scenes where hands are visible. I think maybe 1 in 50 images with visible hands is usable. And when I say usable I don't even mean it looks great. I just mean that the hand isn't so messed up that it's completely unacceptable. Even the "usable" images typically look at least somewhat awkward.
There are more things that are impossible or AI struggles with, but at the end of the day, it just means that many images are just not possible to create. I have to design my videos specifically around the limitations. Even when designing things from the get go around the limitations, I will still run into things that are impossible. In that case I have to either cut that part of the video or come up with an alternative.
It's still possible to create stories though, but the stories have to have larger scale narratives. I have a video I have been working on for a couple weeks called "The Sands of Time" I will be releasing on Friday. The narrative is of planetary scale, so specific people and places aren't needed. I actually have subtitles telling the story along with the images. So narrative pieces are possible but only narratives that are designed around the limitations.
That being said, I have played around with generating AI images and compositing them for animation and other uses. So you could generate images of people, monsters, etc and remove the backgrounds. Pixelmator Pro has some good tools for removing backgrounds automatically for example. You could then generate additional images to use as backgrounds for the scenes. You could then composite scenes for a larger narrative with those pieces. Yes you would have to work around issues with lighting and such, but it is possible. You could even animate them in layers to create kind of a parallax effect between the subjects and the background, which looks pretty cool.
That being said though, especially for realistic images, it can look a little off. That's why I don't do it. I would rather have single images that are fully integrated and feel seamless, even if that creates a lot of limitations. With the right illustrated styles it could look quite good though.
How do you do it? incredible.
I just create a lot of images. :)
Its been so cool watching your channel grow. You're doing a great job!
Thank you. It has been kind of weird, especially since this channel started out as an experiment.