Analysis updated 2026-07-03
Generate Chinese web novel fiction in fantasy or romance style from a short opening prompt.
Run a local browser-based story generator without any cloud dependency using server.py.
Experiment with the RWKV architecture as an alternative to GPT-style text generation models.
| blinkdl/ai-writer | fudan-generative-vision/hallo2 | abhitronix/vidgear | |
|---|---|---|---|
| Stars | 3,698 | 3,698 | 3,697 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | hard | moderate |
| Complexity | 2/5 | 5/5 | 3/5 |
| Audience | general | researcher | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires downloading pretrained model weights separately, GPU recommended for reasonable generation speed.
AI-Writer is a Chinese-language AI tool that generates fictional text, particularly web novel content in fantasy and romance genres. You give it a prompt or opening passage in Chinese, and it continues the story character by character, producing text that follows the style and vocabulary of Chinese web novels. The model was trained on tens of thousands of novels and can generate indefinitely long passages from a short starting context. The tool is built on the author's own model architecture called RWKV, which the README describes as faster than GPT-style models while producing comparable output quality. The model keeps track of the last 512 characters as its working context, using those to predict each new character in sequence. The author includes a plain-language explanation of how this works: each Chinese character is encoded as a set of numbers, the model does a series of mathematical operations on those numbers, and then picks the next character based on probability. The README notes that the project is now outdated and has been superseded by RWKV-Runner, a separate project with larger and more capable models (1.5B, 3B, and 7B parameters). A hosted web version is also available through ModelScope for people who do not want to install anything locally. For those who do want to run it locally, the setup involves downloading pre-trained model weights and running either run.py for command-line generation or server.py for a browser-based interface. It runs on Windows, Linux, and Mac, and can use an NVIDIA GPU for acceleration, an AMD or Intel GPU via DirectML, or just the CPU if no GPU is available. The CPU version is noticeably slower than GPU-accelerated modes. The generated text is described as for entertainment only, and the author notes the model lacks real-world common sense since its training data comes entirely from online fiction.
A Chinese-language AI text generator trained on web novels that continues a story prompt character by character, built on the RWKV architecture, now superseded by RWKV-Runner.
Mainly Python. The stack also includes Python, RWKV.
License information was not mentioned in the explanation.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Verify against the repo before relying on details.