Distillery
Open-source alternative to Midjourney
Start new thread
Tornike Phkhovelishvili

Distillery - Open-source alternative to Midjourney

by
Distillery
β€’
Distillery is the open-source Discord-based text-to-image generator. Built over Stable Diffusion, we use larger AI models to train new models. We are unique with our ultra flexibility, variety of styles and commitment to level up open-source AI community!

Add a comment

Replies

Best
Tornike Phkhovelishvili
Maker
πŸ“Œ
Thanks for your interest in Distillery - the only TRUE open-source generative art service. Our mission is to level up the open source art community. All our methods and models are open sourced. Our images contain in its metadata all information you need to recreate them at home. We are beyond excited to present to PH community our creation: πŸš€ Join us on a journey from AI enthusiasts to one of the robust art makers Distillery is the brainchild of a small but passionate team of AI enthusiasts who embarked on a mission to leverage the power of open-source community. What started as a humble endeavor has now evolved into one of the most sought-after AI generators in the Discord environment. 🌐 Championing the Open-Source Revolution We firmly believe that the future of AI lies in the open-source community. That's why Distillery proudly embraces the open-source ethos, making AI easily accessible to all. We make sure that our learning are spread across the community openly with the hope to make it easier for people to leverage it. 🎨 A Model for Every Occasion Distillery offers four distinct models - Bloody Mary, SDXL, Vodka, and Screwdriver. Each model is tailored for specific use cases, ensuring you get the results you desire. πŸ”₯ Unleash Your Inner Artist with Total Control Our command function empowers you with unparalleled control over the visuals. Play with negative prompts, tweak configurations, adjust aspect ratios, fine-tune denoise strength, and even control the base image. The possibilities are limited only by your imagination! πŸŽ‰ Get Your Creative Juices Flowing with LoRa's With over 80 LoRa's handpicked from open community, we've crafted a style control system that's second to none. It's not just about generating images; it's about having a blast while doing it. πŸ†“ 10 Free Generations Every Day, no restriction with features We believe in giving back to our community. That's why each user enjoys 10 free generations every day. No strings attached! πŸ› οΈ Open Alpha: Your Feedback Matters We're still in open alpha, so we ask for your patience in case of any delays or bugs. Your feedback is invaluable as we work tirelessly to refine Distillery and make it the best it can be.
@tornike_phkhovelishvili Congratulations on the launch, Team Distillery! πŸš€ Your open-source Discord-based text-to-image generator sounds like a game-changer. The commitment to enhancing the open-source AI community is truly commendable. Can't wait to see the amazing images your platform will create. Cheers to flexibility and style variety! 🎨
Tornike Phkhovelishvili
Maker
@michelepomposo Thanks Michele!
Felipe Infante de Castro
Maker
@tornike_phkhovelishvili @michelepomposo Thank you so much, Michele! We aim to live up to your expectations and will work hard to achieve that!
Brian Douglas
@tornike_phkhovelishvili it’s not clear if this is open source or are you using open source? Do you have a link to the source code you can share?
Irakli Pkhovelishvili
Maker
@tornike_phkhovelishvili @bdougieyo that is fair point that we need to address better. In fact it is both, we are using open-source things obviously (we are based on Stable Diffusion) but we have also been trying to contribute back and for every tool, model, etc that we have been creating we share model weights, code, step-by-step guides, datasets, and so on. What's missing is everything to be organized in one GitHub repo since we have been sharing all of these across different channels. The most consolidated is our substack where we do step-by-step style discussions on various topics but mostly on open source model usage and tuning. Here is an example of one of the latest post where we built a small tool https://followfoxai.substack.com... We also have a bunch of things shared on Huggingface https://huggingface.co/pxovelac, Kaggle https://www.kaggle.com/iraklip/d..., a few GitHub places https://github.com/fidecastro/Co... and some of it just in Distillery docs https://distillery-dev.github.io... I think consolidating all of this in a structured way in one place would make it way easier to answer this question and we have added that to our to-do list.
Zeng
Congratulations to you guys! Been playing with Distillery and this is one of my fav AI art generators.
Tornike Phkhovelishvili
Maker
@zeng Thanks Zeng! We love your art on Distillery :)
Irakli Pkhovelishvili
Maker
@zeng, thanks a ton! You have been one of our favorite users so far, and we hope to keep evolving in fun and exciting ways to keep you there!
Zeng
@irakli_ff Thank you so much. That is very nice of you to say that. 😊
John Lim
πŸ”Œ Plugged in
Hi, congratulations on the launch. I wanted to clarify, does 'open source' mean that the code is available for public access? I couldn't locate any repository on the platform.
Tornike Phkhovelishvili
Maker
@john_lim_my Hey John and thanks! Repo is spread out across a bunch of them. We use ComfyUI as the base engine to run Stable Diffusion: https://github.com/comfyanonymou... To use it, we made a connector that lets me edit and send workflows to Comfy using a single line of code. It's gotten 29 stars on github so far: https://github.com/fidecastro/Co... Finally, this is the worker we use for serverless cloud GPUs: https://github.com/fidecastro/di...... (we use a startup called Runpod to host our GPUs) All the models we host are open source; all but one (SDXL) was finetuned by us. This is our main model, Bloody Mary: https://civitai.com/models/11396... Finally, we currently have just over 90 low-rank adaptation ("lora") models in our system, all open source. This is the list of all of them, with the links to download and their respective author for due attribution: https://distillery-dev.github.io...
John Lim
@tornike_phkhovelishvili Thank you for the quick clarification. I wish you the best of luck and success.
Irakli Pkhovelishvili
Maker
@tornike_phkhovelishvili @john_lim_my we also have a blog where we share pretty much all our explorations along with the code, dataset, step-by-step guides, and though process behind it. For example, today, we just added a new feature that turns simple prompts into more fun and interesting ones, and we published a post about it that allows anyone interested to do something similar. https://followfoxai.substack.com...
Ugljeőa Stojanović
Tornike Phkhovelishvili
Maker
@bdougieyo Here is it
Yuki Ogino
Awesome one! Congrats on the launch. I'll give it a try.
Tornike Phkhovelishvili
Maker
@oginokairan Thanks Yuki! We would love to hear your honest feedback
Philipp Shay
Congratulations on launching, and best of luck with your fantastic generative art service! πŸš€ How do the 4 distinct models differ in terms of generating images ?
Tornike Phkhovelishvili
Maker
@kinzarra Thanks Philipp! On high-level difference in type of images they generate. They are all good in respective styles. My co-founder will give a more precise answer on this soon :)
Felipe Infante de Castro
Maker
Hey @kinzarra , Distillery cofounder here. Other than SDXL, which is the original from Stability, the other models are merges that use our finetuned model Vodka as a base (this one https://civitai.com/models/61086...). Bloody Mary is very good as a general-purpose model and in pop culture generations; Screwdriver is the youngest of the models and was created with the intention to bring more photorrealism to the generations, although Bloody Mary keeps surprising us on that. The thing I like the most about Distillery is that we managed to create a very fun way to mix and match loras (low-rank adaptation models). I'd go as far as to say that the models play a secondary role to the loras. Our back-end deals with setting up their weights, so all the user has to do is to write the lora name on the prompt and its activation word.
Matthias Strafinger
Congrats on your Launch. I will definitely give it a try.
Tornike Phkhovelishvili
Maker
@matthias_strafinger Great to hear! Do let us know your feedback.
Irakli Pkhovelishvili
Maker
@matthias_strafinger Thanks! And let us know if you want to achieve something specific - we will try our best to accommodate many use cases.
Ayesha Awan
Congratulations on the launch of Distillery! I am excited to see what this open-source Discord-based text-to-image generator can do!
Tornike Phkhovelishvili
Maker
@ayesha_awan1 Thanks for the support Ayesha!
Irakli Pkhovelishvili
Maker
@ayesha_awan1 Each week or two, we are adding more and more features and options, so I hope we won't disappoint you with what's there to come. Thanks a ton!
Esus πŸ›‘οΈ
Congrats on the launch! Will give it a try.
Tornike Phkhovelishvili
Maker
@esus Thanks Esus! Do let us know your feedback.
Irakli Pkhovelishvili
Maker
@esus Thanks! Give it a try to our new command added today, /suggest will turn your simple prompt into a fun, crazy one!
Yannick
That's another great step to AI mass adoption ! I was curious about why you chose to offer a Discord server rather than a traditional website ?
Felipe Infante de Castro
Maker
@mho22 excellent question, Yannick! We believe Midjourney found a winning UI in Discord. Generative art benefits so much from the social component, and the command line interface is great for rapid prototyping. In short, we wanted to learn by taking the exact same steps they did.
Irakli Pkhovelishvili
Maker
@mho22 Thanks a ton! In fact, we are united around the idea and hope to play our small role in that adoption. Disord is a surprisingly easy option that fits this type of user experience, and we decided to use it as a starting platform. By no means do we intend to stay exclusively on Discord though
Julien Ergan
Congratulations!
Tornike Phkhovelishvili
Maker
@julienergan Thanks Julien!
Irakli Pkhovelishvili
Maker
@julienergan Thanks a ton and we hope to see you on our Discord testing things!
123
β€’β€’β€’
Next
Last