The tool has been designed with a number of optimizations for the characteristics of online literature. On the content acquisition side, the built-in intelligent crawler can automatically recognize the chapter structure of mainstream novel websites (e.g., m.ilwxs.com), and supports breakpoint transmission and retrying of wrong chapters. On the audio production side, the optimized text pre-processing process preserves the dialogue format and narration markers that are unique to web texts.
The commercialization application scenarios contain three directions: 1) individual creators can convert public novels into audiobooks and upload them to platforms such as Himalaya (the project document contains the case of Ximalaya album 88023000); 2) MCN agencies can batch convert popular online articles to build a content matrix; and 3) educational institutions can use it to produce language learning materials. A typical case is to convert a 300-chapter romance novel into a 20-hour multi-character audiobook, which only requires about 15 hours of processing time.
The project adopts the MIT open source protocol, and is allowed to be used commercially but needs to comply with the API usage specification of each speech synthesis platform. Noteworthy technical limitations include: some of the strict protection of the website needs to expand its own crawler module; free version of the TTS model in the continuous generation of 5 hours may trigger the flow-limiting mechanism.
This answer comes from the articleTool to automatically crawl novels and generate multi-character audiobooksThe
English 
简体中文 
日本語 
Deutsch 
Português do Brasil