Unstable Diffusion

Unstable Diffusion is a Discord community and affiliated commercial platform, operated by the company Equilibrium AI, dedicated to artificial intelligence (AI) generated NSFW content built on top of the open source AI image tool Stable Diffusion. ^{[1] [2]} Widely cited as one of the first organized attempts to commercialize AI generated pornography, the project began in August 2022, in the immediate wake of Stability AI's open release of the Stable Diffusion model, and has grown from an informal Reddit thread into one of the most visible adult focused communities in generative AI. ^{[1] [3]} Members share illustrations and AI rendered photographs ranging from softcore to explicit. Many early outputs included grotesque nude figures because of the base model's limited understanding of human anatomy. ^[2]

The community is operated by Equilibrium AI, led by founder and chief executive Arman Chaudhry. ^{[1] [3]} In December 2022, Unstable Diffusion drew international attention when Kickstarter suspended its crowdfunding campaign after pressure from working artists. ^{[3] [4] [5]} The campaign had raised more than $56,000 from 867 backers against an initial goal of $25,000 before being pulled on December 21, 2022. ^{[3] [4] [5]} After the suspension the project pivoted to a self hosted donation channel and, by 2023 and 2024, into a paid subscription website at unstability.ai with custom fine tuned models for adult image generation. ^{[3] [6]}

Overview and key facts

Scroll sideways for more →

How did Unstable Diffusion start?

Unstable Diffusion started in August 2022, almost immediately after Stability AI released the weights for Stable Diffusion under a permissive license. The first hub was a subreddit, which migrated to a Discord server within weeks as Reddit's moderation tools proved inadequate for the volume and explicit nature of the content being shared. ^{[1] [3]}

A server admin told TechCrunch in November 2022 that the goal was "to provide support for people interested in making NSFW," adding: "Because the only community currently working on this is 4chan, we hope to provide a more reasonable community which can actually work with the wider AI community." ^[1] The founders framed the project as an attempt to drag adult AI image generation out of anonymous imageboards and into a moderated space.

The early community split in two directions: sharing AI generated pornography and methods to bypass content filters, and building a custom NSFW image model from open source tools. Although Stable Diffusion shipped with a safety filter intended to suppress nudity and gore, the open source nature of the code made it straightforward for outside developers to disable that filter and retrain on different data. ^[2]

How does Unstable Diffusion work?

Stable Diffusion's training corpus was drawn primarily from the LAION 5B dataset, which contains about 5.85 billion image and text pairs; Stable Diffusion itself was trained mainly on the roughly 2.3 billion English language pairs within it, of which only around 2.9% was NSFW material. ^{[1] [11]} That low ratio explains why nude generations from the stock model often have anatomical errors such as misplaced limbs and distorted hands. ^{[1] [2]}

Unstable Diffusion's response was to assemble its own NSFW dataset and fine tune the base model on it. The team recruited volunteers from the Discord server to label and curate explicit images, an effort that TechCrunch reported had grown to a stated dataset of more than 30 million curated adult images by mid 2023. ^[2] The architecture remains a latent diffusion model: a U-Net denoiser operating on latent image tensors, conditioned by a CLIP text encoder, and decoded by a variational autoencoder. ^[6] What changes is the data and the safety configuration; the default content filter is replaced by an age verification gate, and CLIP based content restrictions are removed during fine tuning. ^[6]

Discord community structure

The Unstable Diffusion server hosts a wide range of AI generated NSFW imagery in different art styles, sexual preferences, and kinks. Coverage from late 2022 catalogued the channels listed below: a men only channel, a softcore "safe for work" stream, channels for hentai and furry artwork, a BDSM and "kinky things" subgroup, a channel dedicated to "nonhuman" nudes, and a genderqueer only channel. ^{[1] [3]}

Members can use the server's Discord bot to generate art and submit favorite outputs to a community "starboard." ^[1] By November 2022 Unstable Diffusion claimed members had generated more than 4,375,000 images through the bot, and the group regularly held competitions "that challenge members to recreate images using the bot, the results of which are used in turn to improve Unstable Diffusion's models." ^[1] These competitions turned the user base into a free reinforcement learning labor pool, with successful prompts and outputs feeding the next training run.

According to Arman Chaudhry, in two months the core team grew to more than thirteen people plus consultants and volunteer moderators. ^[1] He told TechCrunch: "We see the opportunity to make innovations in usability, user experience and expressive power to create tools that professional artists and businesses can benefit from." ^[1]

Patreon and early funding

By November 2022, Unstable Diffusion's Patreon, used to cover compute and development costs, had passed $2,500 per month from several hundred donors. ^[1] The community also secured a five figure cloud hardware grant through an unnamed startup accelerator program. ^[1]

Why did Kickstarter suspend the Unstable Diffusion campaign?

In December 2022 Unstable Diffusion launched a Kickstarter campaign aimed at funding a major training run for a dedicated NSFW model. The pitch promised a custom dataset of approximately 75 million images split into three roughly equal pools: 25 million anime and cosplay images, 25 million artistic images sourced from ArtStation, DeviantArt, and Behance, and 25 million photographic images. ^{[4] [5]}

The campaign raised more than $56,000 from 867 backers against an initial goal of $25,000 before Kickstarter suspended it on December 21, 2022. ^{[3] [4] [5] [10]} Because Kickstarter follows an all or nothing funding model, the pledged money was returned to backers. ^{[3] [4]}

The suspension followed a public outcry from working illustrators. The cartoonist Sarah Andersen, whose comic style had been replicated without consent by users of Stable Diffusion, posted on December 11, 2022: "Shame on @Kickstarter for allowing the Unstable Diffusion crowdfund. You are enabling blatant theft and are funding a tool that can create abusive content such as nonconsensual pornography." ^{[4] [7]} Andersen later joined Kelly McKernan and Karla Ortiz as a named plaintiff in Andersen v. Stability AI, a class action filed on January 13, 2023 in the U.S. District Court for the Northern District of California challenging the use of artists' images to train AI models. ^[7] The fantasy illustrator Greg Rutkowski, whose name was among the most popular search prompts in early Stable Diffusion releases, became another rallying point for artists worried about uncompensated training data. ^[4]

Kickstarter chief executive Everette Taylor said the platform "must, and will always be, on the side of creative work and the humans behind that work." ^{[4] [5]} He listed two policy concerns: whether a project copies or imitates artists' work without consent, and whether it exploits communities or risks harm. ^{[4] [5]} Arman Chaudhry called the suspension a "capitulation," telling reporters: "While Kickstarter's capitulation to a loud subset of artists disappoints us, we and our supporters will not back down from defending the freedom to create." ^[3]

What happened after the Kickstarter suspension?

Within days of the Kickstarter suspension, Unstable Diffusion launched a Stripe powered donation page on its own website. The group reported raising more than $15,000 in the immediate aftermath and roughly $26,000 by mid 2023. ^{[3] [2]} Chaudhry told supporters the team expected continued pushback from Patreon, Kickstarter, and Discord itself, and would build infrastructure that did not depend on any single platform. ^[3]

What is unstability.ai?

The Discord community spun off a paid web platform branded unstability.ai, which sells subscription access to four named in house fine tunes trained on a curated dataset of more than 30 million adult images. ^{[2] [6]} The four headline models address different visual niches:

Scroll sideways for more →

By July 2023, TechCrunch reported that the platform had more than 350,000 users and was generating over 500,000 images per day, run by a core team of five people working full time. ^[2] Those figures are self reported, and later secondary coverage in 2026 continued to cite similar numbers. ^[6] The company frames the platform in expansive terms; one 2023 statement to TechCrunch read: "Our belief is that art, in its many forms, should be uncensored, and this philosophy guides our approach to AI tools and their usage." ^[2] The site offers free, Basic, Premium, and Pro tiers, and reserves commercial use for higher tiers. ^[6]

What are the ethical concerns about Unstable Diffusion?

The ethical concerns around Unstable Diffusion are a sharper version of the debates surrounding generative AI more broadly. TechCrunch summarized the issue this way: "While the use of AI to create porn isn't new, Unstable Diffusion's models are capable of generating higher fidelity examples than most. The generated porn could have negative consequences particularly for marginalized groups, the ethicists say, including the artists and adult actors who make a living creating porn to fulfill customers' fantasies." ^{[2] [3]}

Abhishek Gupta, founder of the Montreal AI Ethics Institute, told TechCrunch that "servers like Unstable Diffusion become hotbeds for accumulating a lot of problematic content in a single place, showing both the capabilities of AI systems to generate this type of content and connecting malicious users with each other." ^{[1] [8]}

The most cited concerns include:

Scroll sideways for more →

A November 2022 essay published by the Montreal AI Ethics Institute argued that the most distinctive risk of systems like Unstable Diffusion is their generative breadth. Unlike traditional deepfakes, which produce a fixed set of altered media, diffusion models can produce a near infinite stream of variations on a given target, defeating hash based filtering and overwhelming moderation teams. ^[8]

Unstable Diffusion's stated policy is to be an "ethical" community: it prohibits child sexual abuse material, deepfakes, and excessive gore. ^[1] Discord users must agree to terms of service and submit images for moderation, and the server runs an automated filter that blocks generations matching a "named persons" database of public figures. ^{[1] [3]} Critics have argued that these measures are reactive and easy to evade given the open ended nature of diffusion outputs. ^{[2] [8]}

Wider context: payment processors and adult AI

The Unstable Diffusion story sits inside a larger pattern of friction between adult AI image generation communities and the platform infrastructure they depend on:

Scroll sideways for more →

Is Unstable Diffusion still active in 2025 and 2026?

Equilibrium AI has funded or been associated with adjacent projects in the space. The most cited example is Waifu Diffusion, a Stable Diffusion fine tune trained on anime artwork that became one of the most widely downloaded community models in late 2022. ^[1] Chaudhry told TechCrunch that he expected Unstable Diffusion to evolve into a broader umbrella organization "supporting other AI powered content generation, sponsoring developers, and providing tools and resources to help with the development of other systems." ^[1]

By 2025 and 2026 the visible center of gravity for uncensored AI image generation had shifted away from any single community. Alternatives include Pony Diffusion V6 XL (a community fine tune of Stable Diffusion XL), FLUX.1 dev from Black Forest Labs, and video focused platforms such as Kling 2.0 and Runway Gen 4. PromptLayer's 2025 overview described Unstable Diffusion as "the uncensored AI Silicon Valley won't touch," reflecting how mainstream platforms and investors have kept their distance. ^[9] A 2026 industry review concluded that "Unstable Diffusion is no longer the center of gravity" in the uncensored AI generation space, although unstability.ai continues to operate with a sizable user base. ^[6] The project still serves as the canonical case study in journalism and academic ethics writing whenever the topic of AI generated pornography and crowdfunding controversies comes up.

Legacy

Unstable Diffusion is studied less for technical achievement than for what its history reveals about the policy environment around open source generative models. The project showed how quickly an open release can be repurposed for uses the original developers did not intend, even when a content filter ships in the box. ^[2] The Kickstarter episode also established an early template for how mainstream crowdfunding platforms would treat AI image projects: by reacting to public outcry from working artists rather than writing detailed rules in advance. ^{[4] [5]}

An unofficial Unstable Diffusion Beginner's guide is available at rentry.co/UnofficialUnstableGuide.

See also

• Stable Diffusion
• Stability AI
• Waifu Diffusion
• Latent diffusion model
• LAION 5B
• Deepfakes
• CivitAI
• AI generated pornography
• Andersen v. Stability AI
• Greg Rutkowski

References

1. Kyle Wiggers, "Meet Unstable Diffusion, the group trying to monetize AI porn generators," TechCrunch, November 17, 2022. https://techcrunch.com/2022/11/17/meet-unstable-diffusion-the-group-trying-to-monetize-ai-porn-generators/ ↩
2. Kyle Wiggers, "As AI porn generators get better, the stakes get higher," TechCrunch, July 21, 2023. https://techcrunch.com/2023/07/21/as-ai-porn-generators-get-better-the-stakes-raise/ ↩
3. Kyle Wiggers, "Kickstarter shut down the campaign for AI porn group Unstable Diffusion amid changing guidelines," TechCrunch, December 21, 2022. https://techcrunch.com/2022/12/21/kickstarter-shut-down-the-campaign-for-ai-porn-group-unstable-diffusion-amid-changing-guidelines/ ↩
4. Luke Plunkett, "Kickstarter Suspends AI-Generated Image (Well, Porn) Project," Kotaku, December 21, 2022. https://kotaku.com/kickstarter-ai-art-image-porn-unstable-diffusion-nsfw-1849921325 ↩
5. Maria Saaikina, "Kickstarter updates its AI art policy following suspension of Unstable Diffusion campaign," Game World Observer, December 22, 2022. https://gameworldobserver.com/2022/12/22/kickstarter-ai-art-policy-suspends-unstable-diffusion-campaign ↩
6. "Unstable Diffusion AI in 2026: Uncensored AI Image Tool," Plisio, April 27, 2026. https://plisio.net/ai/unstable-diffusion-ai ↩
7. "Andersen et al. v. Stability AI Ltd. et al.," class action complaint, Case No. 3:23-cv-00201, U.S. District Court for the Northern District of California, January 13, 2023. ↩
8. Abhishek Gupta, "Unstable Diffusion: Ethical challenges and some ways forward," Montreal AI Ethics Institute, November 14, 2022. https://montrealethics.ai/unstable-diffusion-ethical-challenges-and-some-ways-forward/ ↩
9. "Unstable Diffusion: The Uncensored AI Silicon Valley Won't Touch," PromptLayer Blog, 2025. https://blog.promptlayer.com/unstable-diffusion-the-uncensored-ai-silicon-valley-wont-touch/ ↩
10. GIGAZINE editorial staff, "The project Unstable Diffusion that creates pornographic images with AI is blocked from seeking investment at Kickstarter," GIGAZINE, December 26, 2022. https://gigazine.net/gsc_news/en/20221226-kickstarter-ai-porn-unstable-diffusion/ ↩
11. David Thiel, "Identifying and Eliminating CSAM in Generative ML Training Data and Models," Stanford Internet Observatory, December 20, 2023. https://cyber.fsi.stanford.edu/io/news/investigation-finds-ai-image-generation-models-trained-child-abuse ↩
12. "Civitai Turns to Crypto After Credit Card Processor Ban Over AI Explicit Content," Decrypt, May 2025. https://decrypt.co/322197/civitai-crypto-credit-card-processor-ban-ai-explicit-content ↩