把电子书变成能听的有声书Turning e-books into listenable audiobooks


AI free token/AI免费积分

📦 10 Parts + Conclusion

👉 滑动浏览

PART 01

硬件配置

CPU 就能跑

PART 02

八步制作

从测试到成品

PART 03

离线使用

隐私保护模式

PART 04

踩坑汇总

7 个常见问题

PART ///

写在最后

版权与起步

你有没有过这种想法——手里攒了一堆 EPUB、PDF 或者几万字的 TXT 文件,想在通勤路上、做家务的时候“听”完它们,但一想到要把文件传到某个配音平台上、按字数掏钱,就打了退堂鼓。要是碰到还没出版的书稿、公司的内部材料或者自己写的笔记,上传到别人服务器这件事就更让人不放心了。

Abogen 就是来解决这个问题的。它能让你在自己的电脑上,把电子书直接变成有声书,全程不用联网(除了第一次下载模型),也不用花钱。

这篇文章会手把手带你走完整个过程。我知道你可能是第一次接触这类工具,所以我会尽量讲得直白一些,跳过那些让人头大的术语,让你看完就能动手。

01

PART

Abogen 能干什么

FEATURES · WHAT IT DOES

简单说,Abogen 就是一个“文字转语音”的桌面工具。你给它一本书(电子版),它给你一段音频。目前最新正式版是 v1.3.1,支持以下功能:

支持把 EPUB、PDF、TXT、Markdown 文件转成有声书,字幕文件(SRT、ASS、VTT)也能直接配音。可以按章节或页码选择转换范围,每一章单独保存也可以合并成完整的一本书。

输出格式覆盖 MP3、M4B、WAV、FLAC、OPUS 五种,还能同步生成字幕文件。它支持多音色混合,可以把几个声音按不同比例调出你喜欢的感觉。一次可以排好几本书,按顺序自动处理。

它有 桌面版和网页版两种操作方式。桌面版更稳定,网页版功能更新更快,多了人物识别、多角色配音这些高级功能。

02

PART

我的电脑能跑得动吗

HARDWARE · SYSTEM REQUIREMENTS

普通电脑就能跑,不需要独立显卡。 先转一小段试试速度,心里就有数了。

最基本的条件

1

操作系统:Windows、Linux、macOS 都行

2

Python 版本:3.10、3.11 或 3.12(推荐 3.12,macOS 别用 3.13 会报错)

3

系统依赖:eSpeak NG(免费小工具,装起来很简单)

4

网络:第一次安装和下载模型、音色需要联网,之后不需要

显卡加速对照表

系统NVIDIAAMD无显卡
WindowsCUDA(推荐 12.8)不支持CPU
LinuxCUDAROCm 6.4CPU
macOS--Apple Silicon → MPSIntel → CPU

说白了就是:旧电脑先转一章试听一下,速度能接受就用,不能接受再考虑上显卡。

03

PART

支持哪些文件格式

FORMATS · INPUT & OUTPUT

输入格式方面,桌面版支持拖入或粘贴:.epub .pdf .txt .md .markdown .srt .ass .vtt。网页版支持上传 .txt .pdf .md .markdown .epub,上限 400MB。

输出音频格式有五种,各有侧重:WAV 体积大但无损,适合后期剪辑;FLAC 无损压缩,体积更小;MP3 兼容性最好,手机车机都能播;OPUS 压缩效率高;M4B 能内嵌章节和封面,做完整有声书首选。

中文字幕目前只支持句子级和逗号级的时间轴,没有逐字高亮功能。想要歌词那种跟着朗读一个字一个字亮起来的效果,这套方案暂时做不到。

04

PART

安装教程

INSTALL · STEP BY STEP

不管用什么系统,第一步都是装 eSpeak NG。

方案一:一键脚本(推荐新手用)

先去下载 :https://github.com/espeak-ng/espeak-ng/releases/latest,

找到 .msi 文件下载安装。

然后下载 Abogen 源码压缩包:https://github.com/denizsafak/abogen/archive/refs/heads/main.zip

解压后双击 WINDOWS_INSTALL.bat,脚本会自动装好 Python 3.12 和所有依赖,还会自动检测你有没有 NVIDIA 显卡。装完后程序在 python_embedded\Scripts\abogen.exe,直接双击就能用。

!踩坑提示 🕳

这个脚本会联网下载东西。如果你在公司电脑上操作,可以先看看脚本内容再决定是否运行。

方案二:用 uv 命令行安装

没有 NVIDIA 显卡:

...bash

uv tool install --python 3.12 abogen

有 NVIDIA 显卡(CUDA 12.8):

...bash

uv tool install --python 3.12 "abogen[cuda]" \

  --extra-index-url https://download.pytorch.org/whl/cu128 \

  --index-strategy unsafe-best-match

安装完成后,命令行输入 abogen 就能启动。

Linux 安装

Ubuntu 或 Debian:

...bash

sudo apt update

sudo apt install espeak-ng

uv tool install --python 3.12 abogen

Arch Linux:

...bash

sudo pacman -S espeak-ng

uv tool install --python 3.12 abogen

Fedora:

...bash

sudo dnf install espeak-ng

uv tool install --python 3.12 abogen

装完之后如果命令行找不到 abogen,运行这两行:

...bash

echo "export PATH=\"/home/$USER/.local/bin:\$PATH\"" >> ~/.bashrc

source ~/.bashrc

macOS 安装

...bash

brew install espeak-ng

uv tool install --python 3.12 abogen \

  --with "kokoro @ git+https://github.com/hexgrad/kokoro.git,numpy<2"

注意要用 Python 3.12,别用 3.13,会报错。

05

PART

普通话有声书制作:八步走

AUDIOBOOK · 8-STEP GUIDE

先拿一小段文字跑通流程,确认没问题了再转整本书。 不然等了好几个小时发现音色不喜欢,那才叫崩溃。

STEP 01

准备一小段测试文字

新建一个 TXT 文件,编码选 UTF-8(这个很重要),写入以下测试内容:

...txt

这是普通话语音测试。

重庆银行今天发布了新的公告。

项目版本是二点零,价格是一百二十八元。

这段话能帮你一次性检查:普通话自然度、多音字(“重庆”“银行”)、数字和标点停顿。

STEP 02

启动程序

命令行输入 abogen,或者直接双击桌面图标。把测试 TXT 拖进窗口,也可以直接粘贴文字。

STEP 03

选择语言

这一步很多人搞错。语言一定要选 Mandarin Chinese,Kokoro 对应代码是 z。记住:语言和音色要一起切换,光换音色不改语言,读出来会怪怪的。

STEP 04

挑一个你喜欢的中文音色

目前有 8 个中文音色。女声用 zf_ 开头,男声用 zm_ 开头:

女声:

zf_xiaobei、zf_xiaoni、zf_xiaoxiao、zf_xiaoyi

男声:

zm_yunjian、zm_yunxi、zm_yunxia、zm_yunyang

建议每个音色都点 Voice Preview(试听),重点听:人名地名读得对不对、多音字有没有读错、数字和日期处理得怎么样、中英文混在一起好不好听、标点停顿自不自然。

STEP 05

调整语速

普通话建议从 0.9 到 1.0 这个范围开始试。小说对白可以稍快一点,教材知识类内容最好慢一些。用同一段样音对比着听,选你觉得舒服的。

STEP 06

检查文本内容

转换前先检查这些容易忽略的地方:

1

页眉页脚会不会被反复读出来(很影响听感)

2

目录和页码要不要保留

3

脚注内容会不会打断正文

4

英文缩写需不需要手动改成全称

5

多音字和专有名词要不要提前替换

扫描版 PDF(图片格式的)如果选不中文字,说明没有文字层。Abogen 不带 OCR 功能,需要先用其他 OCR 工具转成可检索的 PDF 或 TXT。

STEP 07

选输出格式

M4B

带章节·完整有声书

WAV

无损·后期剪辑

MP3

兼容性最好

字幕选 Sentence 或 Sentence + Comma

STEP 08

先转一章,满意了再转全书

这是最重要的原则。 先生成一章试听,把多音字、音量、语速、断句、章节标题都确认一遍,没问题了再把整本书加入处理队列。转长书的时候尤其用 CPU 跑的要有耐心,软件支持查看进度、取消任务、逐章保存,中间出问题不会白跑。

06

PART

网页版(Web UI)怎么用

WEB UI · ADVANCED FEATURES

桌面版不够用的时候可以试试网页版。启动命令是:

...bash

abogen-web

然后浏览器打开 http://localhost:8808。

网页版比桌面版多了 人物识别、多角色配音、发音覆盖等高级功能。

!踩坑提示 🕳

这个网页服务默认监听所有网络接口(0.0.0.0:8808),而且没有登录验证。如果你只是自己在本机用,建议把端口限制为本地访问,别把它暴露到公网上。

07

PART

能不能克隆声音

VOICE · MIXER & CLONING

Voice Mixer(音色混合器)可以把几个已有的音色按不同比例混合,调出一种新的声音感觉。但它不能上传真人录音然后训练出一个新声音——没有声音克隆功能。

混合音色的时候也建议先用短句试听。不同组合出来的效果差别可能很大,有的组合甚至会含混不清或者发音不稳定。

08

PART

关于 LLM(大语言模型)

LLM · OPTIONAL ONLY

做普通话有声书完全不需要 LLM。

Abogen 的核心 TTS 引擎是 Kokoro。你导入文档、选择中文音色、生成音频,整个过程不需要 ChatGPT、Claude 或任何大语言模型。

LLM 只在网页版里作为一个可选功能出现,用来做文本规范化,比如处理缩写、撇号和特殊文本替换。但目前的规则明显偏向英语,中文内容开了反而可能帮倒忙。

...env

ABOGEN_LLM_BASE_URL=http://localhost:11434

ABOGEN_LLM_API_KEY=***

ABOGEN_LLM_MODEL=llama3.1:8b

ABOGEN_LLM_TIMEOUT=45

ABOGEN_LLM_CONTEXT_MODE=sentence

只做普通话有声书的话,关了 LLM 更省事,也更保护隐私。

09

PART

如何完全离线使用

OFFLINE · PRIVACY MODE

第一次安装和下载模型、音色需要联网。下载完之后,在设置里点两个按钮:

1

Pre-download models and voices for offline use(提前下载好模型和音色)

2

Disable Kokoro's internet access(禁用 Kokoro 的网络访问)

之后程序就会在完全离线的状态下运行,语音生成和可选的 LLM 处理都留在你电脑本地。

需要注意,以下操作还是会联网:

1

使用远程 OpenAI 兼容 LLM

2

首次下载模型和音色

3

SuperTonic 的首次自动下载

4

自动检查更新

5

Calibre OPDS 和 Audiobookshelf 集成

10

PART

新手常见问题汇总

FAQ · TROUBLESHOOTING

这里把我见过的坑都列出来,你遇到了不用慌:

1

选了中文音色但出来的是英文发音

语言没改过来。记得把语言改成 Mandarin Chinese(代码 z),音色用 zf_ 或 zm_ 开头的。

2

扫描 PDF 转换后没有声音

这种 PDF 是图片格式,里面没有文字。先用 OCR 工具转成可检索的版本。

3

想要歌词那种一个字一个字亮起来的效果

做不到。中文目前只有句子级和逗号级的时间轴,逐词高亮只对英文有效。

4

macOS 装了 Python 3.13 报错

项目要求 Python 小于 3.13,换成 3.12 就行了。

5

网页版上传字幕文件没反应

字幕转语音目前桌面版处理得更好,切回桌面版试试。

6

长书转换要等很久

这是正常的,尤其是用 CPU 跑。先转一章确认设置,再挂那让它慢慢跑。

7

多音字或者人名读错了

先在短样音里发现这些问题,然后用网页版的 发音覆盖 功能修正,或者在转换前直接在文本里替换。

///

LAST

写在最后

LEGAL · FINAL THOUGHTS

隐私方面,如果你开了远程 LLM 规范化功能,你的文本会发到对应的服务器。未出版的书稿、公司材料、个人隐私内容,一定把远程 LLM 关掉,或者改用本机跑 Ollama。

Web UI 的密钥、密码等信息存在本地配置文件里,没有专门加密,所以这个配置目录要当作敏感文件对待,别把它上传到 Git 仓库。

版权方面,Abogen 用的是 MIT 开源协议,Kokoro 用的是 Apache-2.0。但软件开源不等于你转出来的有声书可以随意公开传播。想在公开渠道或商业场景使用有声书之前,先确认原书的版权和音色模型的许可。

如果你的需求很简单——把自己的笔记、买过的电子书或者有授权的材料转成普通话有声书——推荐的起步配置:Python 3.12 + Kokoro + Mandarin Chinese + 任选一个 zf_ 或 zm_ 音色 + 语速 0.9 到 1.0 + 输出 M4B 或 MP3 + 先关掉 LLM。

先跑通一小段,再转整本书。

这八个字能帮你躲掉 90% 的麻烦。

既然看到这里了,如果觉得有用,随手点个赞、在看、转发三连吧。

点赞在看转发

THANKS FOR READING

📦 10 Parts + Conclusion

👉 Slide to browse

PART 01

Hardware Configuration

CPU is sufficient

PART 02

Eight Steps to Production

From Test to Final Product

PART 03

Offline Usage

Privacy-Protected Mode

PART 04

Pitfall Compilation

7 Common Issues

PART ///

Closing Remarks

Copyright and Getting Started

Have you ever had this thought — you're sitting on a pile of EPUBs, PDFs, or tens of thousands of words of TXT files, and you want to “listen” to them while commuting or doing chores, but the moment you think about uploading them to a dubbing platform and paying per character, you back away. And if it's unpublished manuscripts, company materials, or your own notes, the idea of sending them to someone else's server is even more unsettling.

Abogen is here to solve that problem. It lets you turn electronic books into audiobooks right on your own computer, with no need for internet (after the initial model download) and no cost.

This article will guide you through the entire process step-by-step. I know you might be new to tools like this, so I'll keep it simple and skip the confusing jargon — so that after reading, you'll be ready to dive in.

01

PART

What Abogen Can Do

FEATURES · WHAT IT DOES

In short, Abogen is a desktop tool for “text-to-speech”. You feed it a book (electronic version), and it spits out an audio file. The current stable version is v1.3.1, supporting:

Conversion of EPUB, PDF, TXT, Markdown files to audiobooks, and direct dubbing of subtitle files (SRT, ASS, VTT). You can select by chapter or page number, save each chapter separately or merge into a full book.

Output formats include MP3, M4B, WAV, FLAC, and OPUS, with simultaneous subtitle generation. It supports multi-voice mixing — you can blend several voices in different ratios to create your own. You can queue up multiple books for sequential processing.

There are two ways to operate: Desktop and Web. The desktop version is more stable, while the web version gets faster updates and includes advanced features like character recognition and multi-voice dubbing.

02

PART

Can My Computer Handle It?

HARDWARE · SYSTEM REQUIREMENTS

A regular computer is sufficient — you don't need a dedicated graphics card. Just run a short test to check the speed.

Basic requirements:

OS: Windows, Linux, macOS

Python: 3.10, 3.11, or 3.12 (recommended 3.12; avoid 3.13 on macOS)

Dependencies: eSpeak NG (free, easy to install)

Internet: Needed only for initial installation and model/voice download; then fully offline

GPU Acceleration Chart:

The takeaway: Test a chapter first on old hardware. If the speed is acceptable, use it; otherwise, consider upgrading to a GPU.

03

PART

Supported File Formats

FORMATS · INPUT & OUTPUT

Input formats: The desktop version supports drag-and-drop or pasting of .epub, .pdf, .txt, .md, .markdown, .srt, .ass, .vtt. The web version supports uploading .txt, .pdf, .md, .markdown, .epub — up to 400 MB.

Output audio formats: Five options, each with its own purpose: WAV is large and lossless, good for editing; FLAC is compressed lossless; MP3 is most compatible with phones and car systems; OPUS has high compression efficiency; M4B embeds chapters and covers, ideal for full audiobooks.

Subtitles currently support only sentence-level and comma-level time axes — no per-word highlighting. So if you want that karaoke-style word-by-word glow, this setup isn't there yet.

04

PART

Installation Tutorial

INSTALL · STEP BY STEP

No matter the system, the first step is installing eSpeak NG.

Option 1: One-Click Script (recommended for beginners)

Download eSpeak NG from https://github.com/espeak-ng/espeak-ng/releases/latest — find the .msi file and install.

Next, download the Abogen source code zip: https://github.com/denizsafak/abogen/archive/refs/heads/main.zip

Unzip and double-click WINDOWS_INSTALL.bat — it will automatically install Python 3.12 and all dependencies, and check for NVIDIA GPUs. The program is then at python_embedded\Scripts\abogen.exe — just double-click to run.

! Pitfall Tip 🕳

The script downloads files online. If you're on a work computer, review the script before running to ensure compliance with policies.

Option 2: Install via uv command line

Without NVIDIA GPU:

uv tool install --python 3.12 abogen

With NVIDIA GPU (CUDA 12.8):

uv tool install --python 3.12 "abogen[cuda]" \

--extra-index-url https://download.pytorch.org/whl/cu128 \

--index-strategy unsafe-best-match

After installation, type abogen in the command line to start.

Linux Installation

For Ubuntu or Debian:

sudo apt update

sudo apt install espeak-ng

uv tool install --python 3.12 abogen

For Arch Linux:

sudo pacman -S espeak-ng

uv tool install --python 3.12 abogen

For Fedora:

sudo dnf install espeak-ng

uv tool install --python 3.12 abogen

If abogen isn't found after installation, run:

echo "export PATH=\"/home/$USER/.local/bin:\$PATH\"" >> ~/.bashrc

source ~/.bashrc

macOS Installation

brew install espeak-ng

uv tool install --python 3.12 abogen \

--with "kokoro @ git+https://github.com/hexgrad/kokoro.git,numpy<2"

Important: Use Python 3.12, not 3.13, which causes errors.

05

PART

Creating Mandarin Audiobooks: 8-Step Guide

AUDIOBOOK · 8-STEP GUIDE

First, run a small test to validate the setup before converting the entire book — no one wants to wait hours only to find the voice is wrong.

Prepare a test text snippet

Create a new TXT file with UTF-8 encoding (this matters). Write:

Create a new TXT file, choose UTF-8 encoding (this is crucial), and write the following test content:

This is a Mandarin speech test.

Chongqing Bank issued a new announcement today.

This paragraph checks tone, multi-accent words ("重庆" vs "银行"), numbers, and punctuation handling.

Launch the program

Command line: abogen, or double-click the desktop icon. Drag the test TXT into the window, or paste the text directly.

Select language

Crucial step — set Language to Mandarin Chinese (Kokoro code: z). Voice and language must match; changing only the voice won't work properly.

Pick your Chinese voice

There are 8 options: female (zf_) and male (zm_).

Listen to each Voice Preview to check: name pronunciation, tone accuracy, number/date handling, bilingual (C/E) fluency, and punctuation pauses.

Adjust speech rate

Start with ~0.9–1.0 for Mandarin. Dialogue might need faster, educational content slower. A/B test the same sample to find what feels right.

Check text content

Before conversion, verify:

Do headers/footers repeat? (Bad for listening experience)

Should TOC or page numbers be included?

Are footnotes handled correctly?

Are acronyms expanded? (e.g., "USA" → "United States of America")

If it's a scanned PDF and text isn't selectable, you need OCR preprocessing — Abogen doesn't do that.

Choose output format

M4B

With chapters, for full audiobooks

WAV

Lossless, for post-processing

MP3

Most compatible with devices

Subtitles: Use Sentence or Sentence + Comma

Convert one chapter, then the full book

The golden rule: Convert a chapter first to verify settings — tone, volume, speed, breaks, titles. Once satisfied, add the full book to the queue. For long books, especially on CPU, use the progress tracking and cancel/save features wisely.

06

PART

Web UI (Browser) Usage

WEB UI · ADVANCED FEATURES

If the desktop version isn't enough, try the web interface. Run:

abogen-web

Open http://localhost:8808 in your browser.

The web UI adds character recognition, multi-voice dubbing, and pronunciation overrides — features not in the desktop version.

! Pitfall Tip 🕳

The web server listens on all interfaces (0.0.0.0:8808) without login. For security, restrict it to localhost — never expose it to the internet.

07

PART

Voice Cloning & Mixing

VOICE · MIXER & CLONING

The Voice Mixer allows blending existing voices in different ratios to create new tones. However, it does not clone voices — you can't upload a recording and train a new one.

When mixing, test short phrases first. Some combinations may sound unnatural or unstable.

08

PART

On LLMs (Large Language Models)

LLM · OPTIONAL ONLY

You don't need an LLM for Mandarin audiobooks.

Abogen uses Kokoro for TTS. The LLM feature (in the web UI) is optional and used for text normalization — e.g., expanding abbreviations or correcting punctuation. Since the rules were tuned for English, it's often better to disable it for Chinese content to avoid errors.

To turn off remote LLM calls, set these in configuration:

ABOGEN_LLM_BASE_URL=http://localhost:11434

ABOGEN_LLM_API_KEY=***

ABOGEN_LLM_MODEL=llama3.1:8b

ABOGEN_LLM_TIMEOUT=45

ABOGEN_LLM_CONTEXT_MODE=sentence

For pure Mandarin work, skipping LLM is safer for privacy and efficiency.

09

PART

Going Fully Offline

OFFLINE · PRIVACY MODE

After initial setup and model download, enable offline mode:

Settings → Pre-download models and voices

→ Disable Kokoro's internet access

This ensures all processing stays local. Note that some features (like remote LLM, auto-updates, or cloud integration) will fail if offline — so plan accordingly.

10

PART

Troubleshooting & FAQs

FAQ · TROUBLESHOOTING

Common issues and fixes:

Chinese voice returns English pronunciation: Forgot to switch Language to Mandarin (z); voice must match language.

No output for scanned PDF: PDF is image-only; run OCR first to extract text.

Want per-word highlighting like karaoke: Not supported — only sentence/comma time axes are available for Chinese.

macOS Python 3.13 error: Python version >3.13 isn't supported yet — downgrade to 3.12.

Web UI subtitle upload fails: Use desktop version for better SRT/ASS/VTT handling. Long books take forever — normal for CPU; check progress and cancel if needed. Mispronunciations: Use web UI's Pronunciation Override or preprocess text with manual replacements. 11 PART Final Thoughts & Legal Notes LEGAL · FINAL THOUGHTS Privacy: If using remote LLM normalization, your text goes to a third-party server — sensitive data should be processed locally. Web UI settings are stored in plain text — protect that directory. Copyright: Abogen is MIT-licensed, Kokoro is Apache-2.0 — but that doesn't grant you rights to the original books! Always respect copyright laws before distributing or commercializing audiobooks. Recommended starter config: Python 3.12 + Kokoro + Mandarin Chinese + zf_/zm_ voice + 0.9–1.0 speed + M4B/MP3 output + LLM disabled. Test one chapter first, then scale up — it's the fastest way to avoid wasted time. Liked this guide? Please share, like, or follow for more — it helps me write better tutorials!

自动检查更新

Calibre OPDS 和 Audiobookshelf 集成

10

PART

新手常见问题汇总

FAQ · TROUBLESHOOTING

Here I list all the pitfalls I've seen, so you won't panic if you encounter them:

Selected a Chinese voice but got English pronunciation

The language hasn't switched over. Remember to change the language to Mandarin Chinese (code z), and use a voice starting with zf_ or zm_.

Scanned PDF converted to no sound

This PDF is in image format, containing no text. Use an OCR tool first to convert it into a searchable version.

Wanting lyric-like word-by-word highlighting effect

Cannot be achieved. Chinese currently only supports sentence-level and comma-level timelines; per-word highlighting is only effective for English.

macOS 装了 Python 3.13 报错

Project requires Python < 3.13, switch to 3.12.

Web UI doesn't respond when uploading subtitle file

Subtitle-to-speech is currently better handled by the desktop version; switch back to desktop version to try.

长书转换要等很久

This is normal, especially when running on CPU. Convert a chapter first to confirm settings, then leave it to run slowly.

Polyphonic character or name misread

First, identify these issues in short sample sounds, then use the web version's Pronunciation Override to correct, or replace directly in the text before conversion.

///

LAST

写在最后

LEGAL · FINAL THOUGHTS

Privacy-wise, if you turn on the Remote LLM normalization feature, your text will be sent to the respective server. Unpublished manuscripts, company materials, and personal private content should have the remote LLM turned off, or use Ollama locally.

The Web UI's API keys, passwords, etc., are stored in local configuration files without special encryption, so this config directory should be treated as a sensitive file, do not upload to your Git repo.

Copyright-wise, Abogen uses MIT open-source license, and Kokoro uses Apache-2.0. But the software being open-source doesn't mean your generated audiobook can be freely distributed. Before using the audiobook in public channels or commercial scenarios, confirm the copyright of the original book and the license of the voice model.

If your need is simple — convert your notes, purchased e-books, or authorized materials into Mandarin audiobooks — recommended starting config: Python 3.12 + Kokoro + Mandarin Chinese + any zf_ or zm_ voice + speaking rate 0.9 to 1.0 + output M4B or MP3 + turn off LLM first.

Run a short segment first to test, then convert the entire book.

These eight words will help you avoid 90% of troubles.

Since you've read this far, if you found it useful, please share a like, a bookmark, and a forward.

THANKS FOR READING

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry