Lumina Vs Hunyuan 1.1 Vs Pixart Sigma
Added 2024-06-22 19:20:00 +0000 UTC
It's the battle of models with reasonable licenses - Lumina Vs Hunyuan Vs Pixart!
Video link - https://youtu.be/aV_Feh8768M
Which follows prompts best whilst also having a decent style and composition? Well... personally I'm still liking Pixart ;)
Attached below are the workflows used in the video, including the example refiner workflow.
Lumina downloads the models automatically, so super easy there.
Run "pip install flash_attn" in your ComfyUI environment after installing the Lumina custom node, but watch out if you're not using a freedom operating system though, as flash_attn can take a while to build yourself - or you can use the example wheel provided. See the GitHub page for more info:
https://github.com/kijai/ComfyUI-LuminaWrapper
Comments
Thanks for the feedback, and luckily pretty much everything I feature isn't about coding - it's just simple application installs and running stuff that has already been coded :) To cover your numbered concerns: 1. Referring to GitHub pages The GitHub pages provide the most accurate and up-to-date instructions. Sometimes things can change after just an hour or two, so it's always best to go straight to the source for the current info! 2. Step documentation If it's just a single command (like in this case - pip install flash_attn) I will try to include this in the future, though of course this is subject to #1 where things can change. 3. Model Downloads and Directories Again, this information is also typically provided when required. In this specific case, the new Lumina models are downloaded automatically for you. 4. Specific steps to setting up workflows For this I actually do explain everything I do from scratch :) I make notes whenever I'm doing these things, so anything that is off the beaten path I will note down to include. ComfyUI Manager is how I install the custom nodes (where possible). On the rare occasion other things are needed, I will include those too. If anyone hasn't already been through my various basics videos linked in every video description, that should be their first stop! 5. Assumptions of Expertise The assumption of expertise I try to aim for is that individuals will have completed their school courses for ages 5-16, and thus have experience with basic computer usage, file management, python & other langues, can analyze problems, can evaluate and apply new or unfamiliar technologies, etc. If you're outside the UK, then additional information about what is typically taught over here can be found at https://www.gov.uk/government/publications/national-curriculum-in-england-computing-programmes-of-study Overall, the best thing for anyone to do in any particular case, is to provide the specifics of their requirements. Obviously everyone will have the best experience if they're following all of my recommendations and running any python apps in an Anaconda environment on Ubuntu Linux. Of course, I understand people will also want to go their own way and have their own configs and need personalized info specific to them, which is why it helps to provide such specifics in any query.
Nerdy Rodent
2024-06-23 19:28:03 +0000 UTCDear NerdyRodent, I wanted to take a moment to express my admiration for the invaluable resources and knowledge you share on your YouTube channel and Patreon page. Your expertise and contributions to the community are truly appreciated. However, I wanted to bring to your attention a significant issue that many of us, especially those without a strong technical background, have been facing. As someone with limited coding knowledge, I find it incredibly challenging to follow the installation and setup instructions for various applications and tools featured in your videos. The lack of detailed, step-by-step tutorials has been a major barrier. Here are a few specific concerns: 1. Lack of Step-by-Step Instructions: Often, your guides refer us to GitHub pages for more information without providing detailed steps. This leaves us struggling with numerous error messages and no clear direction on how to resolve them. 2. Documentation of Requirements: Including detailed documentation of the various steps, even if access is restricted behind a paywall, would be greatly helpful. 3. Model Downloads and Directories: Instructions on where to download models from and which directories they should be placed in are often missing, leading to further confusion. 4. Complex Interfaces: For example, as OP pointed out, ComfyUI has been particularly difficult to navigate, leaving many of us feeling overwhelmed and discouraged. While I do not expect you to explain everything from scratch for ComfyUI, there are certain steps specific to setting up workflows that require setting up dependencies and models in specific folders. 5. Assumptions of Expertise: It seems that the tutorials assume a level of expertise that many of us do not possess, making the initial setup of new applications or extensions a daunting task. While I understand that your content is immensely beneficial for those with a technical background, it would be incredibly helpful if you could also provide tutorials aimed at non-developers. Detailed, step-by-step guides that cover everything from installation to setup, troubleshooting common errors, and explaining memory requirements would make a huge difference. I have huge respect for the work you do and hope you don't mind me speaking up about this missing chunk in your tutorials. I will continue to seek help and guidance from various sources, but having such comprehensive instructions directly from you would be immensely valuable. I believe that if you can address this issue, your viewer and Patreon base will multiply manifold, as well as the appreciation and gratitude of a larger community stretching beyond the technical experts. Thank you for considering this feedback and for all the fantastic content you create.
TD Film Studio
2024-06-23 11:29:31 +0000 UTCLumina downloads the models automatically, so nothing to do there :) I don't use Microsoft Windows and everyone should always read the instructions, whatever their OS! You are correct that the prompt battle focuses purely on prompts. Lumina downloads the models automatically, so nothing to do there :) I'm not sure what you mean by non-programming people, as there is nothing to do with programming here? Perhaps if you have a specific query?
Nerdy Rodent
2024-06-23 10:51:19 +0000 UTCEverything just goes in the hunyuan dir for the hunyuan node versions. For the "Extra Models" version: mT5 -> t5 CLIP -> clip sdxl-vae-fp16-fix -> vae Hunyuan-DiT -> checkpoints
Nerdy Rodent
2024-06-23 10:46:34 +0000 UTCToo bad you don't bow down to us simple, non-developer users. We need a step-by-step explanation: on your Patreon page, you can do it all by saying "See the GitHub page for more info"... And I get the countless error messages I can't do anything about... Nowhere can I find a word about the memory requirements of the things shown in the video... You do not describe where the models should be downloaded from, to which directory... For me - and maybe I'm not alone - ComfyUI has been a difficult interface so far, and I just hold my head in fright... It's so terribly complicated and does nothing for me but discourage me completely. Please be helpful to non-programming people who just want to play!
Zoltán István Bíró
2024-06-23 06:49:33 +0000 UTCI hate to be the dumb one, but I am still confused regarding which models and files links to download to which directory esp. for Hunyuan. Any guidance would be greatly appreciated. Thanks in advance
TD Film Studio
2024-06-22 20:54:26 +0000 UTC
