With AI having taken a massive boost in 2025 so has the fears and the scaremongering of the process, things like the fake AI images produced by Iran of a downed israeli F-35. And the increase in AI content on youtube and other social media platforms have made it hard to get away from the negative side of it. As someone who is notrious for looking at the negative side of things and working hard now to completely change my mindset I am going to take a look at the postivite side and what AI can do and also talk a little about the grey area. Good vs Bad Fake F35 Crash created by Iranian Goverment to make the west, especially Donald Trump and Israellian President Benjamin Netanyahu & iranians think it had downed one of the 5th generation fighters The Good/Bad debate for AI is one that often crops up in conversation and the like many immediately put every single AI generator in the abusive category its not the case. there is a ton of AI tools that fall into the BAD area that also fall into the GOOD area. its a bit like the story of the dark vs the light dark is seen as bad and light as good I am talking about tools such as Paraphrasers, Meta Generation helpers for websites to help intergration with search engines, Spell checkers, grammar checkers the list goes on and its impossible to list every single one here then we have image generators. To be honest these fall right between the 2 because not every image used to train a AI model such as Flux 2 or Z-Turbo image is copyright material infact alot is stock images from likes of adobe stock, shutterstock and pexels etc is used to do it does that mean its all good? No, far from it its still used for abuse for example the scenario I mentioned at the top and the iranian incident which was designed to ridcule the US, Big Don and israel more than anyone however top photographers, aviation enhausts easily spotted errors in the image and engineering mistakes. There is a lot of content that is copyright and damages creators and that is something that is never okay no matter how you slice it. Does outright complaining make any difference to it. Simple answer in my view is no because all your doing is giving those who do abuse it more reason to keep doing it more and more. If your getting attention for something would you stop? so lets talk about the good side of above. is image generation all bad well actually no in my view if the model is done with stock images that are free to use and follow open rights management then there is no copyright infrigement on the AI model. AI monitoring Before we do our dive into the 2 different methods its important before we start to forget Terminator, UK TV's CBBC kids drama based on books by Gillian Cross's Demon headmaster and series 3's hyperbrain, hollywood fiction. AI does not work like that, "Without human colloboration it is not much smarter than a pencil sharpener". It needs input in order to work its a bit like Johnny 5 in that respect. can it be controlled? answer yes, likes of europe, US, canada, australia and companies such as microsoft, google and samsung are a few who are actively looking at ways to ensure it is controlled and prevent abuse and harm such as revenge porn or adult content being generated on their platforms how they can do this is by controlling the checkpoint in what you can feed to it in the prompts. Image generation This is one part of AI that even screams at and I sort of mentioned it above a little too, in order to look at this part of AI. In my offline test I am using the orignal Flux model from black labs and for the online I will switch to civitAI. So there is 2 forms of running AI generation and I will break down each other down so you can understand it a little more I wont go into tons of detail here and the program I am using for this is a program called Ruined fooocus which itself is a base of the orignal AI generation tool Stable Diffusion Offline This is becoming increasingly popluar as its free, open source, and there is no censorship and on the internet its easy to train new content to work with a base such as Flux which are known as LoRa's. (Low Rank adaptation) So Breaking the first part down simply in order to create a AI image offline you need a Base I.E Flux, Z-Turbo etc & program such as Ruined fooocus & ability to write a basic description for the prompt. I will go over the postivites and negatives at the end of this bit and the tech side and what your going to need to even do all this. Now to keep this simple I wont go into the fine tune settings its dry and it takes practice. So to keep things simple I am going to use a program called Ruined Fooocus and its auto presets which do a pretty impressive job ruined foocus main screen The prompt runied fooocus prompt box in order to create the image we need the prompt the prompt basically is what you need in order to create the image so when I was saying a min ago about AI not being much smarter than a pencil sharpener its not if i create a poor prompt here the result wont be as good " For example if I put cat" then the result is going to create a pretty generic cat however if you spend a bit of time and I do this cinematic film still A tuxedo Maine coon cat outside, snowing, very fluffy cat, blue eyes, snow on cat, surrounded by trees in the forest , shallow depth of field, vignette, highly detailed, high budget Hollywood film, cinemascope, moody, epic, gorgeous,side view you will see the difference this will create. Now that took quite a bit of effort compared to what you will see next with the online method why people still go to the extent of setting up to do the above offline. of course those who are making fake content for youtube often tend to use online creation purely because the they have offset the cost however youtube is clamping down on fake AI video's The tech stuff AI checkpoint's sitting are not small Flux on its smaller variants are still around 12-15GB a piece LoRa's are not controlled Heavy GPU requirements - Flux needs at least an 8GB GPU and a ton of memory on the host (16+ GB) below it will ethier run but take many mins to generate or just crash the system and right now a current generation 8GB is going to set you back at least £300 (400 US) Cost - this can be expensive to do PC components due to the same subject talking about here are not cheap Easy to use - its nearly as easy to use as online based generators with likes of Stable diffusion and Runied FOOOcus Video creation offline takes a lot more memory Inpainting - Adding to images or editing this is actually helpful for creators who make own content but can also be really slow if pc doesnt have the gpu to handle it A bit like the wildwest its lawless as its open source and not controlled Online
