LiblibAI 哩布哩布
China's biggest Stable Diffusion model and LoRA hub
Overview
LiblibAI is China's largest AI image community, 120,000+ shareable models, tens of thousands of LoRAs, cloud LoRA training and ComfyUI-style workflows. Founded May 2023 by ex-Tencent/Alibaba/ByteDance engineers; claimed roughly 10 million creators in its first year.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 4 Sep 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 15,000 monthly compute credits
- 20GB storage
- Commercial-safe models
- Basic editing tools
- 35,000 monthly compute credits
- 50GB storage
- 2,000 accelerated-generation credits
- 50,000 monthly compute credits
- 200GB storage
- More acceleration + task-queue slots
- XL model training support
Use cases
What you can produce with LiblibAI 哩布哩布
- 120,000+ community-shared image and video models
- Tens of thousands of LoRA checkpoints
- Cloud-based LoRA training, including XL models
- ComfyUI-style workflows runnable in-browser, no local install
- API access to in-house Xingliu models and community checkpoints
- Face-swap, product-photo and meme-generation workflow templates
- Text-to-image and image-to-image generation
- 200 free daily compute credits (~20 images)
LiblibAI is Civitai's Chinese-language counterpart, and a bigger one on paper: over 120,000 community models, tens of thousands of LoRAs, cloud LoRA training, and ComfyUI-based workflows for tasks like face-swap or product photography that most consumer tools can't touch. A generous 200 free credits a day covers roughly 20 images before you need to pay.
The catch is trust and language. State broadcaster CCTV tested the app in 2025 and found it would generate explicit content, including a topless dancing video, within minutes of a targeted prompt, bypassing the platform's own review filters; LiblibAI says it patched the gap, but it's a real data point on moderation depth. The interface, community, and most model documentation are Mandarin-only, and output quality swings wildly depending on which community-uploaded checkpoint you pick. Best for technically fluent users comfortable vetting models themselves.
What people say
LiblibAI (哩布哩布AI) is run by Beijing Qidian Xingyu Technology, founded in May 2023 by a team drawing from Tsinghua, Peking University, Carnegie Mellon and alumni of Tencent, Alibaba and ByteDance. Growth was fast and the funding followed: a record-setting Series B in February 2025 took total funding to around US$130 million from investors including Source Code, Shunwei and CMC, and the platform passed 25 million users with roughly 4 million monthly actives by October 2025. The company says it hit close to 10 million creators in its first year, with more than 100,000 original models and over 2.3 billion AI-generated images shared — figures that, if accurate, make it the largest Chinese-language AI image community by a wide margin.
The core pitch is depth in Chinese-specific styles: ink-wash painting, Chinese architectural rendering, donghua-style animation and character designs tuned to local aesthetics that general-purpose Western models handle poorly. Over 100,000 community checkpoints and LoRA adapters can be browsed, downloaded or run in the cloud with no local GPU, and built-in LoRA training lets creators train a personal style model from reference images. LiblibAI also folds in ComfyUI: community workflows run inside its own interface, so face-swap, product relighting and meme pipelines work without installing ComfyUI or learning node graphs. An API calls both the in-house Xingliu (星流) models and custom community checkpoints, including full ComfyUI workflows, and paid tiers carry clearer commercial-licensing terms on covered models.
In 2025 CCTV's testing team found the app would generate a topless dancing video from a targeted prompt within minutes, exposing a gap in content review; LiblibAI responded with a public self-inspection and said it fixed the specific vulnerabilities found.
Membership runs ¥36/month (¥299/year) Basic VIP, ¥59 Professional, and ¥199/month or ¥1,799/year for an Advanced Training tier aimed at people training their own LoRAs. Free users get 200 compute credits a day, roughly 20 generations, and queues slow peak-hour work. Signup, payment and support assume fluent Mandarin and the community runs on WeChat or QQ rather than Discord; an international version exists but there is no meaningful English onboarding path.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including LiblibAI 哩布哩布's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with LiblibAI 哩布哩布 unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to LiblibAI 哩布哩布 directly →
Spotted something out of date? Suggest an update →
Alternatives to LiblibAI 哩布哩布
More in Image Generation
