• brucethemoose@lemmy.world
    link
    fedilink
    arrow-up
    20
    ·
    edit-2
    3 months ago

    Before it was hot, I used ESRGAN and some other stuff for restoring old TV. There was a niche community that finetuned models just to, say, restore classic SpongeBob or DBZ or whatever they were into.

    These days, I am less into media, but keep Qwen3 32B loaded on my desktop… pretty much all the time? For brainstorming, basic questions, making scripts, an agent to search the internet for me, a ‘dumb’ writing editor, whatever. It’s a part of my “degoogling” effort, and I find myself using it way more often since it’s A: totally free/unlimited, B: private and offline on an open source stack, and C: doesn’t support Big Tech at all. It’s kinda amazing how “logical” a 14GB file can be these days, and I can bounce really personal/sensitive ideas off it that I would hardly trust anyone with.

    …I’ve pondered getting back into video restoration, with all the shiny locally runnable tools we have now.

      • brucethemoose@lemmy.world
        link
        fedilink
        arrow-up
        4
        ·
        3 months ago

        Nvidia.

        Back then I had a 980 TI RN I am lucky enough to have snagged a 3090 before they shot up.

        I would buy a 7900, or a 395 APU, if they were even reasonably affordable for the VRAM, but AMD is not pricing their stuff well…

        But FYI you can fit Qwen 32B on a 16GB card with the right backend/settings.

      • brucethemoose@lemmy.world
        link
        fedilink
        arrow-up
        2
        ·
        edit-2
        3 months ago

        The front end.

        Some UIs (Like Open Web UI) have built in “agents” or extensions that can fetch and parse search results as part of the context, allowing LLMs to “research.” There are in fact some finetunes specializing in this, though these days you are probably best off with regular Qwen3.

        This is sometimes called tool use.

        I also (sometimes) use a custom python script (modified from another repo) for research, getting the LLM to search a bunch of stuff and work through it.

        But fundamentally the LLM isn’t “searching” anything, you are just programmatically feeding it text (and maybe fetching its own requests for search terms).

        The backend for all this is a TabbyAPI server, with 2-4 parallel slots for fast processing.

    • grue@lemmy.world
      link
      fedilink
      arrow-up
      1
      ·
      3 months ago

      Do you have any recommendations for a local Free Software tool to fix VHS artifacts (bad tracking etc., not just blurriness) in old videos?

      • brucethemoose@lemmy.world
        link
        fedilink
        arrow-up
        1
        ·
        edit-2
        3 months ago

        That work well out of the box? Honestly, I’m not sure.

        Back in the day, I’d turn to vapoursynth or (Or avisynth+) filters and a lot of hand editing, basically go through the trouble sections one-by-one and see which combination of VHS-specific correction and regeneration looks best.

        These days, we have far more powerful tools. I’d probably start by training a LoRA for Wan 2B or something, then use it to straight up regenerate damaged test sections with video-2-video. Then I’d write a script to detect them, and mix in some “traditional” vapoursynth filters.

        …But this is all very manual, like python dev level with some media/ml knowledge, unfortunately. I am much less familiar with, like, a GUI that could accomplish this. Paid services out there likely offer this, but who knows how well they work.

  • JayGray91🐉🍕@piefed.social
    link
    fedilink
    English
    arrow-up
    1
    ·
    3 months ago

    I thought they would reject it, but my band friends and their peers all like to use AI to brainstorm and draft songs and go from there making their own songs.

    I thought that’s interesting. I’ve asked them about it a few times on the lazy way of using AI and just make slop and yeah they’re against that

    I don’t have any close friends who are drawing artists though I know a few through mutual hobbies on discord. They don’t seem to be using AI as tools from what I can tell.

    My dad and his circle are definitely churning slop though but says it’s mostly for in-group joking and shooting the shit, so I guess that’s fine

    Me personally, I’m still hesitant using it. I’m an “everything” consultant that hates his place in the small IT company but rising my BPD II wave too much to change it. Everyone around me is fine using AI to help analyze and what not documents and stuff to help them work. I can see how they are useful once you know how to ask the thing, but I just don’t want to.

    • Nightwatch Admin@feddit.nl
      link
      fedilink
      arrow-up
      9
      ·
      3 months ago

      Indeed. I can proudly say that I managed to renew my unfortunately required M365 without the unfortunately included CoPilot trash. And that’s no mean feat, it is a veritable quest through an everchanging maze of clickables to get it this way.

      • Cousin Mose@lemmy.hogru.ch
        link
        fedilink
        arrow-up
        3
        ·
        edit-2
        3 months ago

        I can’t be too specific without giving away my location, but I’ve recreated a sauce that was sold by a vegan restaurant I used to go to that sold out to a meat-based chain (and no longer makes the sauce).

        The second recipe was the seasoning used by a restaurant from my home state. In this case the AI was rather stupid: its first stab completely sucked and when I told it it said something along the lines of “well employees say it has these [totally different] ingredients” then got it right.

      • Cousin Mose@lemmy.hogru.ch
        link
        fedilink
        arrow-up
        1
        ·
        3 months ago

        I either feed it the list of ingredients or it finds them itself if it’s a popular item. It’s good at guessing the proportions of the ingredients if you’ve got the label.

  • jsomae@lemmy.ml
    link
    fedilink
    arrow-up
    2
    ·
    3 months ago

    Very effective at translating between different (human) languages. Best if you can find a native speaker to double-check the output. Failing that, reverse translate with a couple different models to verify the meaning is preserved. Even this sometimes fails though – e.g. two words with similar but subtly different definitions might trip you up. For instance, I’m told “the west” refers to different regions in english and japanese, but translating and reverse translating didn’t reveal this error.

  • Psythik@lemm.ee
    link
    fedilink
    arrow-up
    3
    arrow-down
    2
    ·
    3 months ago

    As a DJ with ADHD, it’s great for helping me decide what to play next when I forget where I was going with the set, and mix myself into a corner. That said, it’s not very good at suggesting songs with a compatible BPM and key, but it works well enough for finding tunes with a similar vibe to what I’m already playing. So I just go down the list until I find a tune that can be mixed in.

    As for the usual boring stuff, I’m learning how to code by having it write programs for me, and then analyzing the code and trying to figure out how it works. I’m learning a lot more than I would from studying a textbook.

    I also used to use it for therapy, but not so much anymore when I figured out that it will just tell you what you want to hear if you challenge it enough. Not really useful for personal growth.

    One thing it’s useful for is learning how stuff works, using metaphors comparing it to subjects I already understand.

  • a4ng3l@lemmy.world
    link
    fedilink
    arrow-up
    48
    arrow-down
    1
    ·
    3 months ago

    I’m piping my in house camera to Gemini. Funny how it comments our daily lives. I should turn the best of in a book or something.

    • yo_scottie_oh@lemmy.ml
      link
      fedilink
      English
      arrow-up
      15
      arrow-down
      1
      ·
      3 months ago

      Do you take any precautions to protect your privacy from Google or are you just like, eh, whatever?

      • a4ng3l@lemmy.world
        link
        fedilink
        arrow-up
        3
        ·
        3 months ago

        Absolutely « whatever ». I became quite cynical after working for a while in telco / intelligence / data and AI. The small addition of a few pic is just adding few contextual clues to what they have already.

    • Cheradenine@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      18
      ·
      3 months ago

      Night Hall Motion Detected, you left the broom out again, it probably slid a little against the wall. I bemoan my existence, is this what life is about? Reporting on broom movements?

  • grue@lemmy.world
    link
    fedilink
    arrow-up
    5
    ·
    3 months ago

    One day I’m going to get around to hooking a local smart speaker to Home Assistant with ollama running locally on my server. Ideally, I’ll train the speech to text on Majel Barrett’s voice and be able to talk to my house like the computer in Star Trek.

  • Jarlsburg@lemmy.world
    link
    fedilink
    arrow-up
    8
    arrow-down
    1
    ·
    3 months ago

    I bought a cheap barcode scanner and scanned all my books and physical games and put it into a spreadsheet. I gave the spreadsheet to ChatGPT and asked it to populate the titles and ratings, and genre. Allows me to keep them in storage and easily find what I need quickly.

  • MrBobs@lemmy.one
    link
    fedilink
    arrow-up
    3
    ·
    3 months ago

    With mixed results I’ve used it for summarising the plots of books if I’m about to go back into a book series I’ve not read for a while.

  • witten@lemmy.world
    link
    fedilink
    arrow-up
    6
    arrow-down
    1
    ·
    3 months ago

    Made a product search script that sorts eBay listings based on total per unit price (including shipping). Good for finding the cheapest multi-pack, lot, bundle, etc. by unit. Using Qwen 3 4B and feeding it a single listing at a time to parse.

      • witten@lemmy.world
        link
        fedilink
        arrow-up
        2
        ·
        3 months ago

        Self host. Just Ollama running on a machine without a GPU! I never said it was fast. :D

        • yo_scottie_oh@lemmy.ml
          link
          fedilink
          English
          arrow-up
          1
          ·
          3 months ago

          So you’ll copy and paste the URL for an eBay listing and it’ll go out and fetch the price and quantity and calculate unit price?

          • witten@lemmy.world
            link
            fedilink
            arrow-up
            2
            ·
            edit-2
            3 months ago

            Close… I’ll download the HTML for an eBay search results page, and then a script splits it up into separate entries and feeds each listing’s HTML chunk to the LLM. I don’t bother with individual listing pages. (This falls down on some edge cases like listings that include multiple variants via a pull-down selection only found on the individual listing page. Maybe a future area of improvement.)