THANK YOU Sophus !
and if you aren’t getting sued, you are getting bought out lol
OpenAI’s acquisition of OpenClaw signals the beginning of the end of the ChatGPT era
I just finished watching a video, that suggests we are entering an age of techno feudalism - go on, make a hit of any kind - watch how far you are allowed to let it grow - go ahead make a hollywood hit, you’ll own nothing and love it! lol
1 Like
It can become very overwhelming to try to keep up.
I personally thus far have ever only been interested in the generative AI side, particularly 2D comic and Manga style. images and animation.
I stopped using ChatGPT as soon as grok came out because chatGPT, is actually politically biased in my opinion.
However, it appears that the biggest battle ground will be in the so-called “agentic” area of AI.
With highly specialized chatbots performing tasks for busy people as well as assisting people in coding projects and similar.
I personally have no need for such services.
But I will probably dabble into some vibe coding just for fun at some point if I can think of a need for a little piece of software that I may be able to use in the production of 2D graphic novels and 2D animation.
Who knows?
Interesting times we are in nonetheless. 
1 Like
@AutoDidact Running autonomous AI agents, which can browse the Internet, can also be a a very significant security risk. There are techniques to hijack agents while they scan malicious websites and then these websites can remote control them.
I am well aware of the risks.
Which is why you don’t do any such experimentation on any machine that you actually care about.
I have 5 computers
I would do all of my agentic or vibe coding experimentation over on one of my old laptops or even my second backup lower spec smartphone.
Just would like to thank @Sophus again for that manual.
I am very impressed how good ltx-2 performed on my 3080. Quality prettymuch the same as grok at 720p, though grok gave better results in prompt reading.
On the specific matter of so-called “techno serfdom”, I think the guy in the video is being a bit melodramatic.
Right now, even the poorest people in America are provided with free internet access via free base model smartphones Given away to anyone who is low-income enough to qualify for food assistance medicaid, disability, section 8 housing and other such basic services.
And this is because those basic government services have a preferred primary method of first contact to be made online.
And free public wifi is available in most urban population centers
Also, the notion that the big tech companies will somehow take ownership of all of AI and charge the “serfs’ to use it completely leaves out the free and open source models that many people are already running on their local hardware.
So unless there’s some sort of dramatic change in the fundamental structure of our government.
The big tech companies will never be able to completely eliminate open source software.
Consider that commercial software companies such is Adobe, AutoDesk, Maxon and the like are not, even able to prevent the pirating of their major titles. Just as Hollywood is completely powerless to stop the pirating of the movie and television content from their streaming platforms.
Now those ancient European feudal systems , to which he is comparing our current situation, were based on control of physical distribution models and physical assets , land , tools etc that were not easily obtainable or defendable by the so-called serfs.
This often repeated Mantra of "you will own nothing " seems to assume that it is always desirable or practical for EVERY person to physically own the source of EVERYTHING that they consume, which does not make sense particularly when it comes to something, that’s disposable such as visual entertainment.
1 Like
Meanwhile, I’m just plugging along with my experiments. I will have to post something soon before this thread is running away from me! Many interesting thoughts are being brought up, and I have opinions about those, but that’s not the primary topic of this thread.
1 Like
While debates are raging left and right, I continue to refine my workflow. I have several screenplays I would like to turn into short movies and what I’m showing today is a possible scene for one of them. Many examples that are shown of the latest models feature very fast action, which is impressive but it is also quite forgiving as imperfections are not as noticable.
The scene below shows a young woman sitting at a worktable in her large tent. She is reading a notebook, but she is tired, so she gets up to have a lie-down.
To create this, I rendered the first frame and the last frame in iClone and then used Nano Banana Pro (NBP) to render those in a semi-realistic 3D cartoon style. For the first frame image I specified waves crashing on the beach, which iClone can not do. So I use NBP to change the style but also enhance the image:
Because I am using two images that the AI needs to use to get from the beginning to the end of the clip, it is important that they match, so I spend some time to make sure of that.
Kling 3.0 allows you to specify a character by providing 10 images of the character at different angles, a bit like a LoRA. I found that that really makes a difference, especially for the face when the character is moving. Once you have set that up, you can reuse the character specification in other scenes.
Kling 3.0 can generate audio, but that makes a generation more expensive, so I do that seperately. It also gives me more contro and I actually like working on audio.
In the prompt for the creation of the clip, I specify waves crashing on the beach, which iClone can not do. So besides the animation, the physics you get by using AI is really worth it. For example, at the end, when the girl lies down, the mattress is pushed in and bouncing a bit as it would.
1 Like
This looks like a great way to get consistent AI animated environments using a familiar 3D scene building approach instead of complicated comfy UI nodes.
I think I am going to give this one a try
2 Likes
Wow, that looks good !
There are also developments in the UI of comfy, to make it easy like a Video Editor. (very early stage - Alpha).
I am a heavy ConfyUI user and I like it, it can be complicated that’s true.
The main reason why I use ComfyUI with open source is to train my own Character or motion loras.
You can’t train closed source models like Seedance, Kling, Veo or Sora. You can use reference pictures only and this only very restricted. (when published)
Open source models are about 1y behind the big closed source models but trainable…
1 Like
That is indeed very interesting and I will certainly have a look. It seems a convenient and more direct way to do what I’m tring to do in a hybrid way by creating a 3D scene in iClone first and using the render as input.
1 Like
A little test of the advanced feature of “super Grok” with original Art created at mage.space
1 Like
Looks good. This clealy debunks the idea that you cannot have consistent characters. People just don’t take the time to learn the tools and find the tools that suit them.
I like to add that I also have used Comic Life 3 to create a graphic novel. I found it very east to work with and it makes it easier to make corrections when needed. In my case I used stills from my film Khentopia as basis for panels of my book.
Here is a preview of some pages: Khentopia preview
Currently the book is only available in print. I should probably publish an online version somewhere.
1 Like
One afternoons work
Character created at mage.space
animated with Grok
( which now supports 20 second HD video with
extend from frame option)
1 Like
Latest tests.
Still some minor consistency losses, when parts of the character leaves the frame, but I am getting close to a new 2D animated film making workflow.
Nearly two minutes of story animation in one afternoon.
This is so much fun!! 
1 Like
This really starts to come together. Consistency is indeed something that is improving all the time. And you are able to develop your own style, which is good.
I have been working on something as well, but I got a bit stuck on how to get the outside and the inside of an establishment (a bar in this case) consistent. So I modeled one to use as a reference. However the initial idea came from the AI, including the name:
I started with this interior and then I wanted a corresponding exterior, but that didn’t quite work out:
The problem is that behind the entrance doors you see the street and that is interpreted that way when using this image to generate a video. However, I liked the idea of the orientation of the entrance, so I built a scene in iClone, consisting of the bar and the street:
This evening I created some AI renders with four different Models, using the prompt: “Render as an urban setting with a cyberpunk vibe. It is evening. The street and the pavement are slightly wet and there is a light drizzle.”
Nano Banana 2:
Grok:
Seedream 4.5:
Flux 2 Pro:
I hope this will help with my project. I have been working on realizing one of my scripts but I wanted something playful. So I started with one of my iClone renders of a young woman and asked the AI to have her start walking. So it was interesting to see the scene unfold and her walking in it without me telling what was in it exactly.
Hopefully with my new renders I can develop the “story” a bit further. And like you said, it is a lot of fun.
2 Likes
I wrote this extremely long and detailed reply, but RL won’t let me post it because it has violence death in it. SO, can’t post any of that reply here, but it discussed how I create the story first and then the pics afterwards, as well as the workflow I’m adopting. My experience has been very uneven and that reply detailed the bad parts of my AI journey. The GOOD part that I can show on thread is I was able to take this render done in Iclone:
And with some minor prompts and uploaded images get this:
This image is flawless. It works for the story and I can actually use it. Also it took very minimal effort. I’ll post more as I use it more, but ALL of the pics I do in AI have to fit into the story. Videos are out of the question for me because they just don’t look good (to me). Accuracy is also an issue, and I can expound on that more on this thread IF the images are safe for the thread. Also AI does VFX incredibly well, and those are the scenes I mostly will be using it for. I just can’t post any blood, gore, or violence (which is fine - I didn’t know that beforehand).
1 Like
I’m pleased to see that you started to experiment with this as well. The restrictions in posting images have to do with the fact that the forum has no age restrictions, so there is a filter. It’s also in wording, it get easily flagged even it is common American-English usage.
Although I have created a graphic novel (based on one of my animated movies), I’m foremost a filmmaker so my experiments are aimed at that. I’m also in the RL comfort bubble. as Autodidact would call it, so my aim is to see how far I can get with a hybrid approach: creating scenes in iClone and characters with CC.
iClone doesn’t have much in the way of physics, so having the AI assist in adding that to the scene is a big plus. Simple things like how liquid behaves when someone drinks a glass of wine are easily added with AI, which would be impossible in iClone alone.
This image is flawless. It works for the story and I can actually use it. Also it took very minimal effort. I’ll post more as I use it more, but ALL of the pics I do in AI have to fit into the story.
This is why it’s very important to use an AI System that allows you to create, or upload an image of a character and has a function whereby you reuse that character.
And instruct the AI what is to be done with that character such as changing poses facial expressions even inserting them into different environments.
This is an absolute game changer for creating comics and graphic novels.
And typically, these are only available in the paid tiers of the AI services so while it’s good to experiment with the free services in the beginning.
Ultimately, you will be very limited until you move on to one of the paid tiers.
