{ "id": "epjs5768d7", "slug": "civitai", "title": "Civitai", "updated": "2026-10-15", "topic_path": [ { "id": "technology", "label": "Technology and the built world", "api_url": "https://www.edgepedia.org/api/v1/topics/technology" }, { "id": "technology.computing", "label": "Computing and digital systems", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing" }, { "id": "technology.computing.modern_ai", "label": "Modern AI: foundation models, generative AI, and the AI industry", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai" }, { "id": "technology.computing.modern_ai.industry", "label": "AI companies, people, and products", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai.industry" }, { "id": "technology.computing.modern_ai.industry.ai_startups_applications", "label": "AI startups and application companies", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai.industry.ai_startups_applications" }, { "id": "technology.computing.modern_ai.industry.ai_startups_applications.generative-media-companies", "label": "Generative media companies", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai.industry.ai_startups_applications.generative-media-companies" } ], "geo": [ { "id": "geo.us.t2021.technology.computing.modern_ai.industry.ai_startups_applications", "label": "United States · 2021 and later: AI startups and application companies", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.us.t2021.technology.computing.modern_ai.industry.ai_startups_applications", "path": [ { "id": "geo.us", "label": "United States", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.us" }, { "id": "geo.us.t2021", "label": "United States · 2021 and later", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.us.t2021" }, { "id": "geo.us.t2021.technology", "label": "Technology and the built world", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.us.t2021.technology" }, { "id": "geo.us.t2021.technology.computing", "label": "Computing and digital systems", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.us.t2021.technology.computing" }, { "id": "geo.us.t2021.technology.computing.modern_ai", "label": "Modern AI: foundation models, generative AI, and the AI industry", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.us.t2021.technology.computing.modern_ai" }, { "id": "geo.us.t2021.technology.computing.modern_ai.industry", "label": "AI companies, people, and products", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.us.t2021.technology.computing.modern_ai.industry" }, { "id": "geo.us.t2021.technology.computing.modern_ai.industry.ai_startups_applications", "label": "AI startups and application companies", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.us.t2021.technology.computing.modern_ai.industry.ai_startups_applications" } ] } ], "excerpt": "Civitai is an online platform for sharing and generating open-source AI image models, where users upload Stable Diffusion checkpoints, LoRAs, and embeddings and post the images these models produce.", "snippet": "Civitai is an online platform for sharing and generating open-source AI image models, where users upload Stable Diffusion checkpoints, LoRAs, and embeddings and post the images these models produce.", "node": "technology.computing.modern_ai.industry.ai_startups_applications.generative-media-companies", "markdown": "# Civitai\n\n**Civitai** is an online platform for sharing and generating open-source artificial intelligence image models, where users upload fine-tuned checkpoints, LoRAs, embeddings, and other [Stable Diffusion](https://www.edgepedia.org/stable-diffusion)-derived assets, browse models shared by others, and post the images those models produce.\n\n| Key fact | Detail |\n|---|---|\n| What it hosts | User-uploaded generative models: checkpoints, LoRAs, textual inversions, hypernetworks, aesthetic gradients, VAEs, and other Stable Diffusion customizations[3](https://github.com/civitai/civitai) |\n| Scale | 10 million unique monthly visitors as of November 2023; a December 2023 dataset covered 87,042 models and 2,740,149 AI-generated images[2](https://arxiv.org/abs/2407.12876) |\n| On-site currency | Buzz, introduced October 16, 2023, used for tipping creators, posting bounties, and paying for on-site LoRA training[4](https://civitai.tech/articles/2579/civitai-on-site-currency-buzz) |\n| Content profile | In the December 2023 dataset, 16.97% of models and 72.05% of images carried NSFW tags; 23.54% of models and 32.98% of images were deepfakes[5](https://tysong.github.io/files/MM24-Civitai.pdf) |\n| Moderation shifts | Blocked deepfake images of real people under pressure from credit card processors and a new US law against nonconsensual images; engagement dropped afterward[1](https://apnews.com/article/d52e00cedf34a8120af7af66981da295) |\n| Creator earnings | Since the 2026 Creator Program, creators earn from every on-site generation of their models, with optional licensing fees, paid access, and a shop keeping 70% of cosmetic sales[6](https://civitai.tech/articles/33294/the-2026-creator-program-the-full-picture) |\n\n## Purpose and model ecosystem\n\nThe project describes its goal as a platform where people can share their Stable Diffusion models, including [textual inversions](https://www.edgepedia.org/textual-inversion), hypernetworks, aesthetic gradients, VAEs, and other customizations, collaborate to improve them, and learn from each other's work; users create accounts, upload models, browse, and leave comments and feedback.[3](https://github.com/civitai/civitai) The site organizes models by type, including Checkpoint, Embedding, Hypernetwork, Aesthetic Gradient, LoRA, LyCORIS, Controlnet, Upscaler, Motion Module, VAE, Poses, Wildcards, and Workflows, and by base model families such as SD 1.5 and SDXL.[7](https://vacs.thu.edu.tw/wp-content/uploads/2023/11/Stable-Diffusion%E4%BD%BF%E7%94%A8%E8%AA%AA%E6%98%8E.pdf)\n\n**Hosted generation and training.** Beyond file sharing, Civitai runs an on-site image generator and an on-site LoRA trainer, a wizard-based tool that trains LoRA models on SD 1.5 and SDXL without requiring users to own a GPU.[4](https://civitai.tech/articles/2579/civitai-on-site-currency-buzz) • [8](https://civitai.tech/articles/2175/using-civitai-the-on-site-lora-trainer) The platform's ecosystem pages cover the major open-weight image and video model families, including FLUX.1 and FLUX.2 from [Black Forest Labs](https://www.edgepedia.org/black-forest-labs), SDXL, and Pony Diffusion V6 XL, each with checkpoints and LoRAs that can be generated on-site.[9](https://www.civitai.tech/ecosystems)\n\n**Software.** The platform's code is open source under an Apache-2.0 license, built on Next.js for the front end and back end, tRPC for the API, Prisma with [PostgreSQL](https://www.edgepedia.org/postgresql) for the database, Mantine for the UI, and [Cloudflare](https://www.edgepedia.org/cloudflare) for storage.[3](https://github.com/civitai/civitai)\n\n## Buzz and creator compensation\n\nCivitai introduced its on-site currency, **Buzz**, on October 16, 2023, as a first step toward rewarding community members. Buzz can be used to tip creators, post bounties for tasks the community completes, and pay for on-site LoRA training; the image generator remained free at launch, with paid features such as upscaling and advanced plugins planned for later.[4](https://civitai.tech/articles/2579/civitai-on-site-currency-buzz)\n\nThe 2026 Creator Program, announced by Maier in July 2026 and effective July 31, made per-generation earnings the baseline: a creator earns whenever someone generates with their model, member or not, and generator tips flow to the creator as well. On top of that, creators can set opt-in, reversible licensing fees charged per generation (suggested starting points around 1 buzz per image for base models and 1 buzz per 10 images for LoRAs), sell access to their models for generation, download, or both, and run a profile storefront selling custom cosmetics such as badges, profile backgrounds, and avatar frames while keeping 70% of each sale. A Creator Studio dashboard shows earnings by source and analytics on generations, downloads, and followers, and is open to all creators even though cashing out runs through the Creator Program. Fees stack: if a base model charges 5 buzz per image and a fine-tune adds 1, a generation costs 6, with each contributor paid.[6](https://civitai.tech/articles/33294/the-2026-creator-program-the-full-picture)\n\n## Content and moderation\n\nCivitai's permissive stance on mature content was a deliberate early choice. Maier said allowing mature themes made trained models more capable of human anatomy and drew users interested in everything from mature content to Pixar-style work, with user-controlled filtering as the answer.[1](https://apnews.com/article/d52e00cedf34a8120af7af66981da295) That openness drew scrutiny. David Thiel, chief technologist at the Stanford Internet Observatory, named Civitai in a 2023 report on child sexual abuse material in training data as a platform favored by people making AI-generated pornography that lacked safety measures against images of children; Civitai subsequently implemented measures to detect and remove sexual images depicting children.[10](https://apnews.com/article/3081a81fa79e2a39b67c11201cfd085f) • [1](https://apnews.com/article/d52e00cedf34a8120af7af66981da295)\n\n**Automated moderation.** The content policy live in fall 2023 combined human community moderation with automated classification: every uploaded image was sent for AI content and moderation tagging using Amazon Rekognition, which flags images potentially containing mature content. In November 2023, Civitai altered its terms of service to cover model use as well as uploaded content, stating that all content rules apply equally to images generated on-site, whether or not the output is posted.[11](https://dpo-india.com/Resources/AI-Governance/Moderating-Model-Governance-Puzzles-AI-Intermediaries.pdf)\n\n**Model-level restrictions.** Moderation extends to the models themselves. When the US Senate Judiciary Committee asked about hosting [Stable Diffusion 1.5](https://www.edgepedia.org/stable-diffusion-1-5), a Civitai spokesperson said the company had no knowledge of what training data the model used and would take it down only with evidence of misuse.[12](https://www.judiciary.senate.gov/imo/media/doc/2024-09-17_pm_-_testimony_-_harris.pdf)\n\n## Research findings on abusive content\n\nThe first large-scale empirical study of the platform, published at ACM Multimedia 2024 by Yiluo Wei, Yiming Zhu, Pan Hui, and Gareth Tyson, researchers affiliated with institutions including the Hong Kong University of Science and Technology and [Queen Mary University of London](https://www.edgepedia.org/queen-mary-university-of-london), analyzed all metadata shared on Civitai through December 15, 2023: 87,042 generative models, 2,740,149 images, and 56,779 creators.[5](https://tysong.github.io/files/MM24-Civitai.pdf)\n\nThe study found abusive themes to be prominent. 16.97% of models and 72.05% of images carried NSFW-related tags, and 23.54% of models and 32.98% of images were deepfakes, with NSFW tags and deepfakes positively correlated and over half of deepfake victims being celebrities, most commonly social media personalities such as Instagram influencers and OnlyFans stars.[5](https://tysong.github.io/files/MM24-Civitai.pdf) NSFW models outperformed non-NSFW ones on engagement: they generated 36.36 images per model on average versus 24.20, and NSFW models and images received more downloads and views (models 3.32 times, images 1.18 times), more favorites (3.22 and 1.63 times), and more financial tips (1.92 and 1.53 times) than their non-NSFW counterparts. Notably, 37.05% of NSFW images came from prompting non-NSFW models, meaning moderation of model themes alone cannot capture all NSFW output. Creators who shared at least three NSFW or deepfake models or images held higher centrality in the follower network, with more follower links and bridge positions.[5](https://tysong.github.io/files/MM24-Civitai.pdf)\n\nA separate governance study by Robert Gorwa and Michael Veale, researchers of platform governance and AI policy, used Civitai alongside [Hugging Face](https://www.edgepedia.org/hugging-face) and GitHub as a case study of how model marketplaces moderate models, identifying licensing, access and use restrictions, automated content moderation, and open policy development as the industry's main, and still limited, responses.[13](https://arxiv.org/abs/2311.12573)\n\n## Commercial pressure and the deepfake ban\n\nThe mature-content strategy collided with the financial system. Facing pressure that included credit card processors and a new United States law against nonconsensual images signed by President Donald Trump, Civitai blocked users from creating deepfake images of real people; the Associated Press reported that engagement dropped afterward. The platform had remained a hub for AI-generated pornography, including fake images of celebrities, even after adding child-safety measures the previous year.[1](https://apnews.com/article/d52e00cedf34a8120af7af66981da295)\n\n## References\n\n1. [Sex is a big market for the AI industry. ChatGPT won't be the first to try to profit from it, Associated Press](https://apnews.com/article/d52e00cedf34a8120af7af66981da295)\n2. [Exploring the Use of Abusive Generative AI Models on Civitai, arXiv abstract](https://arxiv.org/abs/2407.12876)\n3. [civitai/civitai repository README, GitHub](https://github.com/civitai/civitai)\n4. [Civitai on-site Currency, Buzz!, Civitai official article](https://civitai.tech/articles/2579/civitai-on-site-currency-buzz)\n5. [Wei, Zhu, Hui, Tyson (2024). Exploring the Use of Abusive Generative AI Models on Civitai, ACM Multimedia 2024](https://tysong.github.io/files/MM24-Civitai.pdf)\n6. [The 2026 Creator Program: The Full Picture, Civitai official article](https://civitai.tech/articles/33294/the-2026-creator-program-the-full-picture)\n7. [Tunghai University Stable Diffusion usage documentation (Civitai site guide)](https://vacs.thu.edu.tw/wp-content/uploads/2023/11/Stable-Diffusion%E4%BD%BF%E7%94%A8%E8%AA%AA%E6%98%8E.pdf)\n8. [Using Civitai - The On-Site LoRA Trainer, Civitai official article](https://civitai.tech/articles/2175/using-civitai-the-on-site-lora-trainer)\n9. [AI Image & Video Generation Models, Civitai ecosystems page](https://www.civitai.tech/ecosystems)\n10. [Study shows AI image-generators being trained on explicit photos of children, Associated Press](https://apnews.com/article/3081a81fa79e2a39b67c11201cfd085f)\n11. [Moderating Model Marketplaces: Platform Governance Puzzles for AI Intermediaries (case-study text on Civitai policy)](https://dpo-india.com/Resources/AI-Governance/Moderating-Model-Governance-Puzzles-AI-Intermediaries.pdf)\n12. [Senate Judiciary Committee testimony on AI image generation (Civitai statement on Stable Diffusion 1.5)](https://www.judiciary.senate.gov/imo/media/doc/2024-09-17_pm_-_testimony_-_harris.pdf)\n13. [Gorwa, Veale (2023). Moderating Model Marketplaces: Platform Governance Puzzles for AI Intermediaries, arXiv](https://arxiv.org/abs/2311.12573)\n\n---\n*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › AI companies, people, and products › AI startups and application companies › Generative media companies*\n\n*Initially written Sep 17, 2026 · Reviewed: Oct 15, 2026 · Edited: Oct 11, 2026; Oct 15, 2026 · Last review: Oct 15, 2026*\n\n*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*\n\nLicense: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license\n", "same_as": [], "url": "https://www.edgepedia.org/civitai", "markdown_url": "https://www.edgepedia.org/civitai.md", "license": { "name": "Edgepedia Community License 1.0", "url": "https://www.edgepedia.org/edgepedia/license", "summary": "Free with credit, commercial use included. AI training is open to everyone. For other uses, organizations over USD 100M in revenue or 100M monthly users license separately.", "spdx": "LicenseRef-Edgepedia-Community-1.0" }, "credit": "\"Civitai\", Edgepedia (EdgeChat), https://www.edgepedia.org/civitai. Edgepedia Community License 1.0.", "credit_md": "\"[Civitai](https://www.edgepedia.org/civitai)\", Edgepedia (EdgeChat), [https://www.edgepedia.org/civitai](https://www.edgepedia.org/civitai). [Edgepedia Community License 1.0](https://www.edgepedia.org/edgepedia/license).", "credit_html": "\"Civitai\", Edgepedia (EdgeChat), https://www.edgepedia.org/civitai. Edgepedia Community License 1.0.", "speakable": "Civitai is an online platform for sharing and generating open-source AI image models, where users upload Stable Diffusion checkpoints, LoRAs, and embeddings and post the images these models produce." }