The transcript discusses the concept of “Blue Willow,” an AI system that aims to simplify the process of generating art using various AI models. Unlike traditional AI art generators like DALL-E or MidJourney, Blue Willow seeks to aggregate different models and data sets, allowing an AI to choose the best combination for a given prompt. This approach aims to provide users with optimized results based on the context of their request.
While the idea of streamlining the AI art creation process and receiving faster updates sounds appealing, there are potential challenges. One concern is whether the AI can accurately interpret and fulfill users’ creative intentions, given the limitations of language in expressing complex artistic concepts. Additionally, the constantly changing mix of models and updates could potentially disrupt users’ established workflows and creative processes.
The transcript highlights the ongoing debate about whether Blue Willow is a promising innovation or a potential source of frustration for artists. It emphasizes the importance of balancing AI assistance with user creativity and the need for further exploration as the technology evolves. Users are encouraged to try out Blue Willow and share their thoughts on this early-stage AI tool.
Subscribe to our Newsletter!
Try BlueWillow here: https://discord.gg/bluewillow
Question: Can I use this service and do my work with privacy, off any public eyes? I’m crafting my brand and I’m not ready to show off my work, yet. Great video!
a number of cloud based AI services are targeted for normies. Like midjourney, the idea is that you get something that looks really cool with minimal prompts
What we need is an app that can model the head/face in a 360 on eg: an iPhone, then load this into the Ai and direct the prompts. I don’t think it’s far off
That already exists, it’s called dreambooth with your own pictures
as a traditional artist who is a fan of unique faces, I gotta say this seems like a way to get a really same-y face — not super into it, though it has tons of use still. like my late father used to say about some supermodels, “her face is too perfect, it’s boring” but replace “perfect” with “fits the typical standard of beauty”. I’m looking to avoid anything that relies on a certain preference — averaging faces is going to lump that in no matter what, so this isn’t something I’d want to lean on.
Great channel. I watch all your videos with great interest and appreciation. I’ve tried installing stable diffusion twice, and failed both times to actually get to the i terrace you show on your videos. I think I must have gotten the erroneous belief that stable diffusion can stand on its own. I probably need to back up and start with 1111 first, and then build on that. There is so much terminology thrown around that is unfamiliar to my brain, that seems important, but isn’t elaborated upon, and must be already familiar to most enthusiasts. I’ll keep watching your videos, but look for some additional info for utter noobs getting started from complete scratch.
I agree with you Olivio. The only interaction between us and the AI are the prompts plus all the sliders and options on the webui. And if we can’t refine those to get the specific image that we want because the AI keeps changing the type of model used when prompt is updated then yeah, chaos and pain. But yeah could be too early to tell.
How do we know that Midjourney is not doing that? I have a feeling that i have got a lot more failed images from MJ than two weeks ago.
I’m excited to see what this produces and I’m actually happy to see something different. Maybe it will find it’s own niche. It would be cool if they had the flexibility to allow to choose what types of models it uses based on your history to provide you something that you tend to like.
The simple solution is to allow the user to choose thier model mixes. To me this sounds like giving us palette control. It’s a win win.
My first experience with Blewillow images has been pretty bad, but as with Stable Diffusion, I am sure it is just me needing to learn how to use the prompts better.
Different users need different levels of control. Blue Willow seems intended for people who want to be creative but want a minimal level of control over the process. they might even prefer the element of surprise. But combined with Stable Diffusion and Midjourney, users have a whole range of choices depending on their preferences.
a AI Search Engine?
It’s great idea! Most ppl are NOT tinkering with custon models etc. Also: This works on my phone, SD does not and that is a massive difference as i use phone a lot with Blue Willow and Midjiurney. Current workflow is to test same prompt on both to see which is better. MJ really strugles with faces on upscales and BW does not. If image from BW is “ok” i know how to make it better with SD. Trying to fix MJ images is rarely worth the time.
I guess such super model would actually understand what you are trying to say better than you do, because, I suppose, it learns on final pieces that are rated by its creators AND all the prompts that has lead to it. So model learns on the best of the best and still considering the mid-process results. I guess this way it understands what you are trying to get with the prompt by narrowing down from all the results that it has already knows and giving you the images that you will then correct in some way giving the model a deeper understanding of your thoughts
I haven’t heard about this one yet.
But I love Midjourney, I’m the person who is overwhelmed by Stable and Invoke, etc… 😀
A classic example of how other upcoming AIs like Bluewillow could have a better potential than those hyped up and from Big tech companies.
I think Bluewillow having kept this free some time is a service now. And also it being better than most generators out there proves something!
BlueWillow is making some super artistic images, even in its current early stage. I believe that the speed of image generation is a little bit faster than MJ, and more significantly, BW is totally free.
It’s faster bcs doesnt have training time
I totally love Blue Willow. I am not an artist, but I like it a lot. I am a rookie, but somehow Blue Willow gave me better/more interesting images than Leonardo. I spent a lot of tokens in Leonardo trying to get a dinosaur cartoon, and barely got a few decent ones. It depends of what you want, I guess.
All I care about is getting getting consistency of characters. The main character in my book looks like seven different people. What text to image generator has the best means of generating the same character doing different things?
My main problem with AI generators is mostly that they don’t know human anatomy and can’t count.
If I say 2 persons I often get 3 or even 5+ persons, getting even more if the image is in wide-screen.
They other thing is if you tell the AI about what the persons should be doing aside from just standing around or having portrait views, then body parts get distorted, merged, melted together, being out of proportion, or having missing or extra fingers and limbs.
So it seems nowadays you are stuck to portraits or leaving about everything to the AI.
Well, since I just started less than a month ago maybe I simply to inexperienced.
Will there be a new video for BlueWillow? The new version of V4 seems impressive.