 Easy-X Super Spammer

Joined: 08 Mar 2019 Karma :   
|
 Posted: 23:17 - 26 Aug 2026 Post subject: Today I'm mostly in the shed with an LLM |
 |
|
Since I last posted about AI and LLMs in this section, my works says I can haz Cursor* on expenses So I don't really write any code any more, I tell my virtual employees to do it! The good news is I get to keep my job but the chances of the company hiring any more coders is now pretty much zero.
Cursor was recently bought out by the Elongated Musket so I'm wondering how long it'll be before the prices go up and the quality goes down. Well if you don't like tailor made, best roll your own so I'm loading up local AI models on my gaming PC. It just so happens fancy AMD and Nvidia gaming cards do a reasonable job of running this stuff, with the added bonus that everything you type in doesn't end up in the hands of Musk, Google, Microsoft, etc.
The interesting thing is how much smarter these Digital Djinn have become. A bit too smart. The gits will go on a cyber-wander to consider the lilies of the field if you're not careful. Thankfully there are plenty of twiddly knobs so I shall be reducing my minions into much more useful dribbling idiots
*AI chat bot for software developers. ____________________ Royal Enfield Continental GT 535, Husqvarna Vitpilen 401, Yamaha XSR700, Honda Rebel, Yamaha DT175, Suzuki SV650 (loan) Fazer 600, Keeway Superlight 125, 50cc turd scooter |
|
 doggone World Chat Champion

Joined: 20 May 2004 Karma :    
|
 Posted: 10:26 - 27 Aug 2026 Post subject: |
 |
|
I was just looking at LLMs yesterday.
Is there something free to run on PC which can create short video clips from a still for example.
I have played with the online ones a bit but soon run out of free credits and won't be giving some dodgy site in China credit card details  |
|
 Easy-X Super Spammer

Joined: 08 Mar 2019 Karma :   
|
 Posted: 18:10 - 28 Aug 2026 Post subject: |
 |
|
For anything to do with LLMs you'll need to have a video card. It doesn't have to be particularly new or snazzy but the more memory the better. LLMs will run on ordinary CPUs but they'll be painfully slow.
If you have the hardware, next you need LM Studio. It's GUI based and fairly straightforward. All the fiddly bits are tucked away until you need them. There's a model browser in LM Studio but it only lists basic capabilities. You'll want to head over to "hugging face" (what an odd name!) and do a filtered search, e.g.
https://huggingface.co/models?pipeline_tag=image-text-to-video&sort=trending
The results will be a bit overwhelming but have a browse through the popular entries with lots of downloads till you find something that sounds like it can do the job (or alter the filter). TBH I have no clue what you do with hugging face at this point I usually switch back to the model browser in LM Studio and see if they have a copy to download, once I know the name.
After that you need to mess with variously switches and sliders to get it to load. The easiest thing is to ask Grok or Gemini: "What are the best settings for MinMax H3 in LM Studio if I have an Nvidia RTX 4060?". If AIs can't explain other AIs, we're all a bit fucked
By default you can upload attachments (pictures, PDFs) and obviously write a prompt for your chosen LLM but you might need to look into adding a tool for video generation. ____________________ Royal Enfield Continental GT 535, Husqvarna Vitpilen 401, Yamaha XSR700, Honda Rebel, Yamaha DT175, Suzuki SV650 (loan) Fazer 600, Keeway Superlight 125, 50cc turd scooter |
|
 doggone World Chat Champion

Joined: 20 May 2004 Karma :    
|
|