Amuse AI Demo: Stable Diffusion on AMD GPU Made Easy! (No WebUI)

Поділитися
Вставка
  • Опубліковано 31 гру 2024

КОМЕНТАРІ • 62

  • @HugoCNobre
    @HugoCNobre  4 місяці тому +2

    What's your favourite local GenAI model for images? Let me know! As always all links will be in the description 👆

    • @HugoCNobre
      @HugoCNobre  4 місяці тому +2

      Leave your questions down below!

    • @TheGarfield1959
      @TheGarfield1959 2 місяці тому +1

      yo i keep getting a error that goes : index was out of range. must be non negative and less than the size of the collection. (parameter 'index') help me please i cannot find ANY docummentation about it...

    • @TheGarfield1959
      @TheGarfield1959 2 місяці тому +1

      im trying to use the vid2vid mode

    • @TheGarfield1959
      @TheGarfield1959 2 місяці тому +2

      I found the issue, my video wasnt encoded properly and showed 0 fps. I re encoded it at 60 fps and it works slowly but finely on my rx580 8gb.

  • @CapaUno1322
    @CapaUno1322 4 місяці тому +6

    Seriously you just install it and it runs on AMD? Like wow! That's so cool! ;D

    • @HugoCNobre
      @HugoCNobre  4 місяці тому +5

      Yeah, it just works. However it still has fewer models, but the performance on the existing ones is really good.

  • @MasBayu48
    @MasBayu48 3 місяці тому +8

    why amuse 2.0 tutorial so rare on youtube .

    • @HugoCNobre
      @HugoCNobre  3 місяці тому +3

      Don't know exactly why. But it was one of the reasons I created the video (low competition keyword with lack of content)... besides the fact that I was already using the app beacuse I have an AMD System.

  • @rodopil1161
    @rodopil1161 2 місяці тому +1

    Merci beaucoup ! thanks a lot , so useful and well presented !!!

    • @HugoCNobre
      @HugoCNobre  2 місяці тому +1

      Your welcome, thanks!

  • @xgx7999
    @xgx7999 День тому +1

    Potentially useful tool, gimped by build-in censorship mechanics, turned up to hilariously absurd levels. Tried it to generate images which were intended to be used as a visual aid for a short presentation about Classic Greece art and architecture; input materials included (among others) few pictures of real Greek sculptures of that era which depicted naked subjects, because that how some artist from 2500 years ago create them. All pictures generated based on these were blurred into complete uselessness. I'd rather use something that give me more 'hassle' but treats me like an adult person and won't decide above my head what kind of content I'm allowed to create.

    • @HugoCNobre
      @HugoCNobre  День тому

      Yeah, probably not a good option for NSFW content. Try other models with a Web UI maybe.

  • @alanluna3011
    @alanluna3011 Місяць тому +1

    Amazing.

  • @RuslanZhizherin
    @RuslanZhizherin 3 місяці тому +1

    Finally I can use NPU at my Ryzen 8845H! It works, and quite fast 🔥🔥🔥

    • @HugoCNobre
      @HugoCNobre  3 місяці тому +1

      Yes it does however, version 2.0 worked better for me than 2.1.0 in terms of performance and stability (I'm using a GPU).

  • @CapaUno1322
    @CapaUno1322 4 місяці тому +2

    Thanks man! You earned yourself a sub....so can you use the models from CIVITAI and all that jazz?

    • @HugoCNobre
      @HugoCNobre  4 місяці тому +3

      Thanks! As I understand it... not currently. However most of those from CIVITAI are not ONNX, which is what Amuse Ai uses when installing from their Model Manager. Maybe they will add the ability to import safetensors in the future or just convert them.

  • @sculptallday6960
    @sculptallday6960 3 місяці тому +2

    How to add weight to words? I tried both brackets and pluses and it doesn't work

    • @HugoCNobre
      @HugoCNobre  3 місяці тому +1

      I don't think that works on Amuse Ai.

  • @sirchristopher7684
    @sirchristopher7684 3 місяці тому +3

    Do you know if you can train your own art style to achieve a specific look?

    • @HugoCNobre
      @HugoCNobre  3 місяці тому +2

      I don't think so. That's not what the app is built to do. However, there might be other projects for that like ComfyUI. But what you are referring to sounds a lot like creating an actual Stable Diffusion model because there’s only so much you can do at the prompt level.

  • @Gray-Today
    @Gray-Today 7 днів тому +1

    If I like this software, I'll buy an AMD GPU just to run it.

    • @HugoCNobre
      @HugoCNobre  3 години тому

      You don't have to, but you can if an AMD GPU is what fits you better. NVIDIA also works well with AI using other apps and Web UIs.

  • @Soiypakit0
    @Soiypakit0 23 дні тому +1

    puedo usar flux?

    • @HugoCNobre
      @HugoCNobre  23 дні тому +1

      Sí, puedes instalar flux.1-schnell con 1 clic desde el "model manager" u otros modelos de flux usando flux "model install template".

  • @atomicskull6405
    @atomicskull6405 4 місяці тому +2

    Does it work with Pony XL V6?

    • @HugoCNobre
      @HugoCNobre  4 місяці тому +1

      Not sure, never tested that model. It's easier to just install ONNX models. I'm not sure about using safetensors in Amuse ai.
      community.amd.com/t5/ai/amuse-2-0-models-import/td-p/698593

  • @INFO-DIY
    @INFO-DIY 2 місяці тому +1

    MY PC => onnxruntime failed to initialize amuse. HELP ME

    • @HugoCNobre
      @HugoCNobre  2 місяці тому

      I have other types of issues, depending on the amuse version. After closing and opening the app again it was working. I don't really know how to solve this specific issue. If you have AMD hardware and windows 11 amuse is supposed to work out of the box.

    • @INFO-DIY
      @INFO-DIY 2 місяці тому +1

      @@HugoCNobre I am using windows 10, gpu rx 7600 and RAM 16 GB.. Should I upgrade my windows to 11?

    • @HugoCNobre
      @HugoCNobre  2 місяці тому +1

      @@INFO-DIY I don't recommend upgrading to windows 11 if your current version is working well for you just because of Amuse AI. You could try a Stable Diffusion Web UI, it might work better.

  • @deadhand6873
    @deadhand6873 Місяць тому +1

    Can ryzen 7 8845hs run amuse with no problems?

    • @HugoCNobre
      @HugoCNobre  Місяць тому +1

      It should run fine, but runs better if you have a dedicated GPU. Only way to know for sure is to test it.

  • @chealay5744
    @chealay5744 3 місяці тому

    Can I use the model that I already downloaded from Civitai?

    • @HugoCNobre
      @HugoCNobre  3 місяці тому

      No, I don't think you can use safetensor files with Amuse AI, I have tried it before didn't work. It must be a ONNX Model. Some of the models on Civitai do have a version that works on Amuse AI but it's usually on huggingface. However 2.1.0 has more models available with 1-click installs. You migh want to use a WebUI for those type of models if you can't find an alternative 😄

  • @theGreaterAwareness
    @theGreaterAwareness 3 місяці тому +1

    Spoiler alert, any of the inpainting in Amuse does not work unless you have Windows 11. I just bought the 7900XTX and gave my kids my 2080 Super thinking I was doing myself a favor - turns out the 2080 Super actually does inpainting while the 7900XTX results in Amuse being the go-to tool. Not only is it an alpha level graphical user interface, it doesn't work on Windows 10 64 bit. I thought I was buying hardware when I bought the 7900XTX, not software. Does anyone know why Windows 11 will work and not Windows 10? I can't see any significant difference other than Mr Gateway wants some more money to fund a virus so we can replace all our public holidays with "Happy Virus day!". The 7900XTX should generate me a windows 11 with the size of it, I had to buy a new PC case, it's like fitting a vacuum cleaner into a lunchbox. One good thing is that the GPU is more quiet than my PC fans but that probably explains why it does a whole lot of nothing without Windows 11.
    I specifically bought the AMD 7900XTX so I could use Amuse. I'm amazed that it doesn't work on Windows 10 64, it's not like things have changed much, make a Windows 10 compatible version or give windows 11 as a courtesy when people spend $1500 on a GPU that doesn't work without windows 11. I get better generative AI from my 2080 Super as it supports CUDA. But it seems when it comes to CUDA, AMD is CUDA SHUDA WUDA. So Using Windows 10, I try to do anything in design and it results in an error. My first time with an AMD GPU, it feels like I just got diagnosed with AIDS from a cheap GPU instead of realizing that at least if I buy a nerfed NVIDIA the AIDS has a tradeoff to counter balance the lack of support, documentation etc. I mean - at least give me poor performance but an error so that I am forced to upgrade to windows 11. Money don't grow on trees, it's summoned into reality by corrupt people that press a button and make themselves money and then it all flows into how banks make money out of money they don't own - there you go, I pretty much explained inflation while trying to point out that Amuse is useless without windows 11.

  • @wagnerfreitas3261
    @wagnerfreitas3261 4 місяці тому +2

    valeu porra, gostei

  • @Jaksary
    @Jaksary 4 місяці тому +2

    Damn this only works on windows 11 and I'm not going to install that. I guess I'll just switch to linux.

    • @HugoCNobre
      @HugoCNobre  4 місяці тому +1

      Yeah, only on Windows. For a Linux native solution you can look at a Web UI and run ONNX Models (for similar performance).
      Here you can find some Web UIs for Stable Diffusion:
      www.reddit.com/r/StableDiffusion/wiki/local
      These should also work:
      Automatic1111 AMD Fork by lshqqytiger
      ComfyUI

    • @Jaksary
      @Jaksary 4 місяці тому +1

      @@HugoCNobre yeah I'm running win10 now.

    • @Jaksary
      @Jaksary 4 місяці тому +2

      @@HugoCNobre it's weird they don't say it's windows 11 only, it just says "windows machine"

    • @HugoCNobre
      @HugoCNobre  4 місяці тому +1

      ​@@Jaksary Being and AMD collab, there's a chance it might come to Linux later.

    • @Jaksary
      @Jaksary 4 місяці тому +1

      @@HugoCNobre I think there are better solutions for linux :)

  • @758185luan
    @758185luan 3 місяці тому +1

    My pictures are all blurry, can you help me?

    • @sandwichninja
      @sandwichninja 2 місяці тому +2

      That's the censorship feature to block adult content. It's incredibly restrictive even if you're not trying to make lewd images.

    • @sandwichninja
      @sandwichninja 2 місяці тому +2

      I found how to turn off the content filter.
      Just navigate to "C:\Program Files\Amuse\Plugins\ContentFilter", and in the "ContentFilter" folder delete the contents, or simply move it elsewhere.
      There are other layers of censorship in AMUSE AI at the moment. For instance, certain explicit prompts are blocked, but removing the contents of that folder will clear up most of the issues.

    • @758185luan
      @758185luan 2 місяці тому +1

      @@sandwichninja thank my friend

    • @sandwichninja
      @sandwichninja 2 місяці тому

      @@758185luan You're welcome!

  • @giuseppeciccotti7714
    @giuseppeciccotti7714 3 місяці тому +1

    better use automatic with zluda

  • @vsc208
    @vsc208 3 місяці тому +1

    Tem canal em portugues?

  • @sandwichninja
    @sandwichninja 2 місяці тому +1

    AMUSE AI has way too much censorship. This diminishes its usefulness a great deal.

    • @HugoCNobre
      @HugoCNobre  2 місяці тому

      I don't know if Amuse AI has censorship or if that's something to do with the models. Not everyone needs to create NSFW images. You can always try another stable diffusion Web GUI.

    • @sandwichninja
      @sandwichninja 2 місяці тому +1

      ​@@HugoCNobre Seems like I need to unpack a few things for you.
      First of all AMUSE AI absolutely has a mature content filter independent of the individual models. for those who want to remove it they can navigate to the "content filter" folder located here: C:\Program Files\Amuse\Plugins\ContentFilter and simply delete its contents. There are other layers of filtering such as certain blocked prompts, but this content filter is the main one and removing it clears up most of the issue.
      Secondly, the reason why this is such a huge issue is that many of the available models are heavily trained on lewd images and gravitate towards those types of results even when the user isn't attempting to make them. This combined with the ultra-strict content filter means that even pictures that aren't lewd or even remotely risqué often get blurred.
      Thirdly, the blur that people are complaining about is applied _after_ the image is generated, so changing around prompts doesn't help very much. With some models you can't even type the words "woman", "female", "girl", etc. without AMUSE blocking the result. So for instance, an actual female user can't even make a female character avatar for her social media. Obviously this isn't the intended result.
      Lastly, absolutely *_no one_* watching this video is able to optimally use other SD frontends. We're AMD users trying AMUSE as a last resort.

    • @sandwichninja
      @sandwichninja 2 місяці тому

      @@HugoCNobre Did you delete my comment? Here it is again:
      Seems like I need to unpack a few things for you.
      First of all AMUSE AI absolutely has a mature content filter independent of the individual models. for those who want to remove it they can navigate to the "content filter" folder located here: C:\Program Files\Amuse\Plugins\ContentFilter and simply delete its contents. There are other layers of filtering such as certain blocked prompts, but this content filter is the main one and removing it clears up most of the issue.
      Secondly, the reason why this is such a huge issue is that many of the available models are heavily trained on lewd images and gravitate towards those types of results even when the user isn't attempting to make them. This combined with the ultra-strict content filter means that even pictures that aren't lewd or even remotely risqué often get blurred.
      Thirdly, the blur that people are complaining about is applied after the image is generated, so changing around prompts doesn't help very much. With some models you can't even type the words "woman", "female", "girl", etc. without AMUSE blocking the result. So for instance, an actual female user can't even make a female character avatar for her social media. Obviously this isn't the intended result.
      Lastly, absolutely no one watching this video is able to optimally use other SD frontends. We're AMD users trying AMUSE as a last resort.