- Civitai Newsletter
- Posts
- 5 minute Rapid LoRA Training on Civitai!, Natural Language Auto-Captioning, Promising New Base Model revealed & More!
5 minute Rapid LoRA Training on Civitai!, Natural Language Auto-Captioning, Promising New Base Model revealed & More!
Greetings, Civitans! Welcome to the fortieth edition of Civitai Biweekly, your essential guide to all recent developments and updates within Civitai.
This week, we’ll cover some extremely exciting new Civitai & Generative AI developments, including…
Huge LoRA Trainer Captioning Update! 🧾
Flux LoRA Training - 60x speed boost! ⏩
GGUF File Format Support! 💾
Welcoming new Artists in Residence! 🖌️
Introducing a Game-Changing New Base Model! 🚀
…and more!
Civitai Developments
1️⃣ LoRA Trainer Natural Language Captioning
We’re excited to announce the release of Natural Language Tagging in the Civitai LoRA Trainer, powered by the immensely popular JoyCaption model. This feature allows you to effortlessly generate long-form, natural language sentences that describe your data set image contents with incredible accuracy. These detailed captions are the preferred method for Flux LoRA training, providing the best training results!
2️⃣ Five minute LoRA training?
You read that right! We're now offering a lightning-fast Rapid Flux Training option in the Civitai on-site Trainer. Thanks to some arcane optimization magic, you can now train a Flux LoRA in under five minutes - without sacrificing quality or fidelity! And here’s the best part: to celebrate this monumental development, we’re offering 50% off Rapid Flux Training Buzz costs until Tuesday, Sept 17, 2024!
See our Civitai LoRA Trainer Guide to get started.
3️⃣ GGUF Format Added
Flux may have steep system requirements, but our clever community has found a way to run it on mid-range hardware using techniques more commonly seen in Large Language Models. With GGUF Quantization methods, Flux’s size and requirements are cut in half!
We’re excited to announce that Civitai now supports the upload of models with the .gguf file extension.
Contests
💫 Flux Training Contest 💫
The Civitai Flux Training Contest has officially wrapped up, with tens of thousands of fantastic submissions! Our team are still hard at work processing the results - it’s been an incredibly tough competition to judge, with so many outstanding entries! Stay tuned for on-site announcements, as the winners will be announced very soon!
As always, a huge thank you to everyone who participated! All entrants should have received their stunning participation profile cosmetics by now - apply them with pride! | Flux Training Contest Profile Badge |
Civitai AiR
Studio Update Thirteen
We’re thrilled to welcome four amazing artists to the residency program in Studio Update Thirteen! Here’s a sneak peek at their unique projects:
🎬 Christopher Manzione is reimagining 2001: A Space Odyssey frame by frame using AI, blending photographic eras from 1826 to 2000 with cutting-edge generative models.
🌍 Riccardo Arena explores speculative narratives that merge human geography and biological mysticism, producing visual stories like Geograficzny Okulistyczne and WALKA-ULUṞU that blur the lines between myth and science.
🎥 Andre Oshea is diving into perspectival anisotropy, generating AI videos that reflect how the same events produce entirely different interpretations based on the observer’s context.
🗿 Rachel Jackson is exploring false indexicality, using AI to create synthetic artifacts around the story of Burrows Cave, challenging the boundaries between history and fabrication.
Check the full update for more on these incredible artists and projects!
Generative AI Updates
New Base Model - Ideogram 2.0!
Ideogram have just launched their latest model, Ideogram 2.0, which is billed as a significant advancement in text-to-image generation. According to Ideogram, the model offers industry-leading capabilities for realism, graphic design, typography, and more. Built from scratch, 2.0 is said to outperform other models in multiple areas, including text rendering accuracy - watch out Flux!
The model introduces a variety of style options, supports custom aspect ratios, and includes features like enhanced photorealism, and color palette control. Ideogram positions 2.0 as a game-changer for designers and content creators, while maintaining competitive pricing in the AI image generation space, and we’re working to make it available via the Civitai on-site Generator!
AI Image & Video’s New Horizon Event
Join us on October 10th for an exciting event during #SFTechWeek, where a panel of industry leaders will discuss the booming AI Image, Video, and Audio landscape and how to build a company in this dynamic space. Featuring insights from Yishu Miao (Haiper.ai), Sabba Keynejad (VEED.io), our own Justin Maier and Jose Florido (Freepik), this event will also showcase major announcements, including Haiper 2.0's SOTA Video Model and Civitai's Project Odyssey AI Film Competition - Season 2! |
Twitch News
New Streams & Captivating Guests!
We hosted our first ever Civitai Leadership Team Stream!
Co-founders Justin and Max joined Team Leaders Tyler, Matty and Ally for a recap of accomplishments from the previous quarter, and to share insights into what’s coming soon for Civitai.com!
If you missed the stream, catch the replay on the Civitai YouTube Channel now!
Civitai Education
Updated Guides & New Education Hub Features
Flux Guide Updated - New Details!
We’ve added exciting new information and resources to our Flux Quickstart Guide, including how to use new GGUF Quantized models, how best to train for likeness capture, and how to leverage ControlNets and IP Adapters with Flux!
We’ve also provided (simple!) pre-built ComfyUI workflows to help you get started!
LoRA Trainer Guide Updated
We’ve updated our LoRA Trainer guide with all the details you need on Natural Language Captioning, and the phenomenal new Rapid Flux Training mode!
Whether you're just beginning your Generative AI Art journey or you're already a seasoned pro, the Civitai Education Hub is your go-to destination for top-tier educational content - https://education.civitai.com
Resource Showcase
Text Logo - Style LoRA for Flux
Creator ImJohnJohn refines Flux’s (already awesome!) textual capabilities, with their Text Logo Style LoRA. We love to see LoRAs specifically targeting Flux’s text-generating ability, pushing the boundaries of image generation!
You can start using Flux on Civitai’s Image Generator or with your local generation interfaces. Don’t miss out - get involved now!
Minimalist Vector Art - LoRA for Flux, SDXL, Pony
Creator ArsMachina, leaves no Prompter behind with a stunning set of minimalist style LoRAs for Flux, SDXL, and Pony! Whichever your favorite model; Minimalist Vector Art has you covered.
Civitai in StatsCivitai stats covering the previous two weeks! |
Models Uploaded: 15k
Images & Videos Uploaded: 2.1M
Reviews: 781k
Comments: 37k
Image Reactions: 11.2M
New Users: 152k
Support Tickets Resolved: 918
Number of meetings where we almost agreed on something: 12
Thanks for sticking with us all the way to the end! Here’s your well-earned Newsletter Buzz ⚡ - consider it a ‘thank you’ for being a valued, informed, member of the Civitai community!
Streaming Schedule
Want to connect directly with the Civitai team, learn more about our platform, or gain insights from leading experts in Generative AI?
📅 Check out our Twitch streaming schedule and mark your calendar! In addition to our regularly scheduled Office Hours and AI Video streams, we have some thrilling, must-see events lined up that you won't want to miss!
KFTIGER LoRA Training Stream - September 20th.
Civitai AiR Roundtable #2 - September 27th.
We look forward to seeing you there!