LiblibAI: All-in-One AI Art and Video Creation Platform from China 312

What is Liblibai?

LiblibAI is a one-stop AI painting and video creation platform launched by Beijing Qidian Xingyu Technology. Starting out as a community hosting Stable Diffusion models, it has gradually evolved into a multi-modal AI professional creation studio covering images, video, 3D, LoRA training, and more.

LiblibAI has now gathered over 20 million creators, with a model library exceeding 100,000 entries and an average of more than 5 million images generated per day. The 2.0 version released in October 2025 significantly enhanced its video generation capabilities. In July 2025, it launched Xingliu AI for design scenarios, and in March 2026 it released the video creation platform LibTV.

Why Choose LiblibAI

LiblibAI is one of the AI creation communities with the most model resources in China. With a single account, you can access mainstream image and video models without recharging across multiple platforms; over 100,000 LoRA style models are searchable and ready to use; from a minimalist generator to ComfyUI workflows, both beginners and experienced users can find an operating depth that suits them.

Core Features

• Three operation interfaces: Minimalist Generator (for beginners), WebUI and ComfyUI (for professional parameter tuning)

• Image Generation: Integrates models such as Seedream 4.5, Qwen-Image, Midjourney V7, FLUX.2, Nano Banana, supporting text-to-image and image-to-image, and can be paired with 100,000+ LoRA style models

• Video Generation: Integrates Kling 2.6, PixVerse V5.5, Hailuo 2.3, Vidu Q2, Tongyi Wanxiang 2.6, supporting text-to-video and image-to-video

• Effects Library: 500+ templates, including human transformation, old photo restoration, building growth, etc. Upload materials and apply directly

• Online LoRA Training: Provides preset modes for portraits, art styles, ACG, etc., with no local deployment required

• Post-production tools: HD repair, smart image expansion, one-click matting, partial redraw, facial repair, prompt reverse inference

• LibTV Agent (launched July 2026): Over 80% success rate for one-shot generation of videos under 3 minutes

Use Cases

• Self-media: Xiaohongshu covers, WeChat public account illustrations, short video materials

• E-commerce Operations: Product scene images, marketing creative images

• Design Professionals: Illustrations, posters, architectural renderings

• Covers over 20 fields including photography, anime, e-commerce, Guochao, 3D, etc.

• General Users: Avatars, wallpapers

• Short Drama Teams, Advertising Companies: Use LibTV for commercial ads and brand short films

Pricing

• Free Users: Claim compute credits daily by logging in; standard resolution single image costs about 1 credit

• Basic VIP: 36 yuan/month or 299 yuan/year, includes 15,000 compute credits, 20GB cloud storage, 300 accelerations, 2 parallel tasks

• Training Premium Member: 199 yuan/month or 1,799 yuan/year, includes 50,000 credits, 200GB storage, supports XL model training

• Compute Packs: Pay per use, suitable for temporary large-volume image generation

• Xingliu AI Membership: Starting from about 59 yuan/month

Frequently Asked Questions

Q1: Is the free version enough?

A: You can claim free compute credits every day, which is generally sufficient for ordinary use. If you create frequently, need to train models, or produce 4K large images, you should consider subscribing to a membership.

Q2: Can the generated images be used commercially?

A: Yes, but it depends on the specific model and material licenses. Before formal commercial use, it is recommended to carefully review the platform's license terms, especially for client projects and ad placements. The platform itself has blockchain certification and commercial license agreement generation features.

Q3: Can Chinese prompts be used?

A: Yes, the platform has built-in automatic translation, and its Chinese comprehension is better than most overseas platforms.

Q4: Can I train my own LoRA?

A: Yes, just upload images and the system will automatically tag them. It is recommended to prepare at least 200 high-quality images for stable results.

Q5: What export formats are supported?

A: PSD layered files, vector graphics, 3D models, etc. are all supported, and can be connected to downstream scenarios such as printing, animation, and VR.

Q6: Will there be queues during peak hours?

A: Free users may have to wait about 40 seconds between 8 PM and 10 PM. VIPs have acceleration privileges, which can significantly shorten wait times.

Q7: How long does model training take?

LoRA training generally takes anywhere from tens of minutes to a few hours, depending on the amount of material and queue conditions, with no need to stay at the computer.

How to Use LiblibAI

• Visit the official website to register an account; supports login via phone number, WeChat, or QQ

• On the homepage, select "Image Generation" or "Video Generation". In Minimalist mode, choose a model and enter prompts to generate images

• Prompt Writing: Combine "subject + scene + style + lighting + image details", e.g., "high-end perfume bottle on a black marble table, cinematic lighting, luxury ad photography"

• Reproduce Others' Results: Go to the model image replication area, select a favorite image, and click "Generate Online" on the right to automatically copy the model, prompt, and parameters

• Advanced Parameter Tuning (WebUI/ComfyUI): Choose the DPM++ 2M Karras sampler, set iteration steps to 30~40, set CLIP skip to 2, and choose VAE 840000

• Video Creation: Go to "Video Effects" or LibTV, select a template and upload materials, or use Agent to generate a complete short film with one sentence

How Does LiblibAI Compare with Competitors

Related News:

LiblibAI 【Chao Plan】 Original Model Submission — Win a 5,000 RMB Prize!

LiblibAI is back online! Filing may be completed

LiblibAI x BlueFocus "Double Festival Celebration" Creative Poster Challenge