Civitai

October 2026

1. Oct 29
   New feature

   Civitai launched Storylines, where users start a story and anyone can continue from any chapter with their own post, branching the story.

   What this means for you: Storylines lets you start a story on Civitai and lets anyone add the next chapter. Each chapter is a post. If you continue from the end of any chapter with your own post, the story branches into a new path.

   How to use it: Storylines is live on Civitai now. Start a story, or open any chapter and continue it with a post of your own. More details are in the article at https://civitai.com/articles/37518.

   I'm about to start a branching story on Civitai's Storylines, where each chapter is a post and anyone can continue from the end of any chapter. Help me write an opening chapter for a story about [topic or genre], and suggest three different directions other people could take it next. My style is [describe tone].

   Copy for your agent

   @hellocivitai on X

2. Oct 21
   New model

   Civitai added Krea 3 Ultra, Krea's highest thinking mode image model, for users to generate with on the site.

   What this means for you: Civitai now lets you generate images with Krea 3 Ultra, the highest thinking mode of Krea's image model. Krea says it ranks #4 on Arena's multi-image editing leaderboard.

   Why it matters: It comes six days after Civitai added Krea 3 on Oct 15, which edits images from one sentence using up to 6 reference images and outputs at 2K.

   How to use it: Open the Civitai generate page with the Krea 3 Ultra link in the post (civitai.com/generate, with the model set to 2k resolution and thinking mode Ultra) and start generating.

   I use Civitai, which just added Krea 3 Ultra, an image model with a high thinking mode that works with multiple reference images. Help me write three clear prompts for it. I want to make: [describe image]. My reference images show: [describe references].

   Copy for your agent

   @hellocivitai on X

3. Oct 19
   New feature

   Civitai added a Turbo mode for MOSS Image that uses 8 steps and costs about a quarter of the Buzz per image.

   What this means for you: Civitai now has a Turbo mode for MOSS Image, an image model. It is one checkbox that uses 8 steps (the number of passes the model makes to draw a picture) and costs about a quarter of the Buzz, Civitai's on-site credits, per image.

   Why it matters: It comes the same day Civitai added MOSS Image itself on Oct 19.

   How to use it: Tick the Turbo checkbox when generating with MOSS Image on Civitai. Details are in the article at civitai.com/articles/36843.

   I generate images on Civitai and just learned MOSS Image has a Turbo mode with one checkbox, 8 steps and about a quarter of the Buzz per image. Help me plan a testing session: which kinds of prompts to try first and how to compare Turbo with normal mode. My Buzz budget: [amount].

   Copy for your agent

   @hellocivitai on X

4. Oct 19
   New model

   Civitai added MOSS Image, a 5B open-weights text-to-image model under Apache 2.0, and users can train LoRAs on it in the Training Studio.

   What this means for you: Civitai now offers MOSS Image, a text-to-image model with open weights, meaning the model files are public. It is built to render text that reads correctly, to follow a JSON layout you give it, and to accept prompts up to 4K tokens (roughly 3,000 words). You can also train a LoRA, a small add-on that teaches a model your own style or subject, on it.

   How to use it: MOSS Image is live on Civitai. To train a LoRA, use the Training Studio. Details are in the article at civitai.com/articles/36843.

   I use Civitai and just learned it added MOSS Image, an open-weights text-to-image model that renders readable text and accepts JSON layout control. Help me write three prompts that test its text and layout. What I want to make: [poster, logo or other image, plus the exact words].

   Copy for your agent

   @hellocivitai on X

5. Oct 15
   New model

   Civitai added Krea 3, an image model that edits from a single sentence using up to 6 reference images, with 2K output.

   What this means for you: Krea 3 lets you change a picture by typing one sentence, and you can give it up to 6 reference images to guide the result. It keeps characters looking the same from image to image and produces images up to 2K resolution. It has Instant and Medium thinking modes.

   Why it matters: It comes two days after Civitai added HeyGen Video on Oct 13, so the site now covers both new image and new video models.

   How to use it: Krea 3 is live on Civitai now. Civitai has an article with details at civitai.com/articles/36718.

   I just learned Krea 3, now on Civitai, can edit an image from a single sentence using up to 6 reference images, and keeps characters consistent. Help me write a clear one-sentence edit instruction and choose reference images for this project: [describe your image and the change you want].

   Copy for your agent

   @hellocivitai on X

6. Oct 13
   New model

   Civitai added HeyGen Video, a general-purpose text, image or reference to video model with sound, at 50% off through October.

   What this means for you: Civitai, a community library of image and video AI models, now offers HeyGen Video. It is HeyGen's first general-purpose video model. It makes a video clip from text, from an image, or from a reference, and every clip comes with sound.

   How to use it: HeyGen Video is live on Civitai now, at 50% off through October. Details are at civitai.com/articles/36653.

   Civitai now offers HeyGen Video, which makes video with sound from text, an image, or a reference. Help me write 3 prompts for a short clip. The clip is for [describe purpose] and should show [describe scene]. For each, say what sound or audio I should describe.

   Copy for your agent

   @hellocivitai on X