
Midnight Radio
Velvet · narrator
● Narrator · suspense / contemporary
Warm and close, made for late-night thrillers and love stories.
Most people looking for novel to audiobook software have tried a phone's read-aloud feature and found the problem: one voice reads every line, so by chapter three nobody can tell who is talking. With AI Drama Voice you format a chapter as "Name: line", paste it in, and the narration goes to a narrator while every character gets a voice of their own, with per-line emotion. You download that chapter's MP3 and SRT, then do the next chapter with the same cast sheet.
Updated Nov 5, 2026Fact-checked Nov 5, 2026
Chinese · The Call at 3 A.M.
Chinese · Script excerpt · 3 characters

Xu Yan (older sister)Whisper
念念……是你吗?
Niannian… is that you?

Xu Nian (younger sister)Excited
你是谁?这个号码,是我姐姐的!
Who is this? This is my sister's number!

Xu Yan (older sister)Choked up
念念,真的是你……三年了,我每天晚上都打这个号码,从来没有人接过。
Niannian, it really is you… Three years. I've called this number every night, and no one ever answered.

Xu Nian (younger sister)
你到底在哪儿?
Where are you?
This is the real dubbing studio with an audiobook sample loaded. Swap in a scene from your own novel and the cast list and line-by-line tags appear right away. This step runs in your browser and costs no credits.
Narrator
Midnight Radio · 4 lines
3 roles · 7 lines
561 / 8,000 charsSame cast · Chinese
A phone call from a suspense novel: a hushed narrator, an older sister on the edge of tears, a younger sister growing more frightened. Both clips are the same scene in two languages, here to show how a full-cast audiobook differs from a single reader (third-party placeholder audio, to be replaced before launch).
Script excerpt · Chinese · The Call at 3 A.M.

Xu Yan (older sister)Whisper
念念……是你吗?
Niannian… is that you?

Xu Nian (younger sister)Excited
你是谁?这个号码,是我姐姐的!
Who is this? This is my sister's number!

Xu Yan (older sister)Choked up
念念,真的是你……三年了,我每天晚上都打这个号码,从来没有人接过。
Niannian, it really is you… Three years. I've called this number every night, and no one ever answered.

Xu Nian (younger sister)
你到底在哪儿?
Where are you?

Xu Yan (older sister)Whisper
我在家。在你的房间里,坐在你的床上。
I'm home. In your room, sitting on your bed.

Xu Nian (younger sister)Whisper
不可能……我现在就躺在床上。
That's impossible… I'm lying in bed right now.

NarratorCalm
许念下意识地看向床边。月光底下,床单平平整整,什么都没有。
Xu Nian glanced at the edge of the bed. In the moonlight the sheet lay perfectly flat. Nothing was there.
Script excerpt · English · The Call at 3 A.M.

NarratorCalm
On the seventh ring, Xu Nian answered. At first there was only static on the line, a steady hiss, like heavy rain.

Xu Yan (older sister)Whisper
Niannian… is that you?

Xu Nian (younger sister)Excited
Who is this? This is my sister's number!

Xu Yan (older sister)Choked up
Niannian, it really is you… Three years. I've called this number every single night, and no one ever picked up.

Xu Nian (younger sister)
Where are you right now?

Xu Yan (older sister)Whisper
I'm home. In your room. Sitting on your bed.
For novel to audiobook work, don't paste the whole book at once. Working chapter by chapter means every chapter can be checked and redone on its own, and it matches how audiobook platforms take uploads: one episode at a time.
Step 1
Split the book into one file per chapter and strip author notes, chapter ads, garbled characters and repeated headings. If a chapter is very long, split it again at a scene break.
Step 2
Leave narrative paragraphs as they are; they go to the narrator. Put each line of dialogue on its own row as "Xu Nian: Where are you?". Write inner thoughts as "Xu Nian (thinking): …" and put emotions in brackets, such as "Xu Yan (choked): …".
Tip · Find-and-replace on quotation marks gets you most of the way; then read through once to fix the speakers.
Step 3
After the first chapter is analysed, note down "character → voice": which voice narrates, which voices the leads and the main supporting roles get. Spell every name the same way throughout the book and use the sheet for every chapter.
Step 4
Paste the formatted chapter into the dubbing studio or upload it as a .txt file. Check that each character's voice matches your sheet, tick English or other languages if you want them, then start generating.
Step 5
Download the chapter's MP3 and SRT and file them as "Title - Chapter N - language". Listen through once; if only a few lines are off, edit those lines in the studio and regenerate just them.
In an audiobook the narrator matters most: it stays with the listener for the whole book. Choose the narrator first, then give the leads and recurring characters voices that are easy to tell apart.

Velvet · narrator
● Narrator · suspense / contemporary
Warm and close, made for late-night thrillers and love stories.

Classic · folk tale
● Narrator · historical / martial arts
A storyteller's rise and fall, good for ensemble casts.

Mature · poised
● Narrator · first-person heroine
Mature and composed, common for novels told by the heroine.

Soft · soothing
● Narrator · cosy / romance
Gentle and unhurried, easy to listen to for hours.

Deep · commanding
● Male lead · commanding
Low and imposing, for CEO and power-play leads.

Bright · playful
● Heroine / younger sister
Bright and young, clearly apart from a mature narrator.

Warm · grounded
● Elder · father / mentor
Deep and steady, for parents and the person behind the twist.

Clear · energetic
● Young hero
Clear and driven, for fantasy and coming-of-age leads.
Rows without a speaker go to the narrator; lines labelled with a name go to that character's voice, so listeners know who is talking without seeing the text.
The voice library is sorted into narrators, male leads, heroines, elders and villains; with a cast sheet, no character changes voice over dozens of chapters.
Words such as "choked", "whispering" or "sneering" in brackets become emotion tags that apply to that line, so a suspense scene doesn't sound like a news read.
Save a formatted chapter as .txt and upload it; both UTF-8 and GBK-encoded Chinese text are read correctly, so web-novel files don't turn into garbage.
Once the Chinese chapter is done, tick English, Japanese, Korean and more; the same cast voices them all and each language exports its own MP3 and SRT.
There are four common novel to audiobook routes, and they differ mainly in two things: can they do multiple characters, and can the result be published. For your own commute, a phone's read-aloud is enough. For an audiobook you want to publish, you need a cast, steady output and the rights.
| Route | How it works | Multiple voices | Best for |
|---|---|---|---|
| Phone or reading-app read-aloud | Open the e-book and press play | Usually one voice throughout | Personal listening, not publishing |
| Open-source TTS tools | Run them locally or on a server and script the chapter splitting and audio joining yourself | Depends on your own setup and scripts | People who code and like to tinker |
| Record it yourself and upload | Record yourself or hire narrators, edit, then upload | One narrator doing voices, or a team | Premium titles that need human performance |
| AI Drama Voice | Format as "Name: line" and generate MP3 + SRT per chapter | Voices assigned automatically, kept fixed by your cast sheet | Long novels with many characters, foreign editions |
A Chinese web novel of 100,000 characters usually runs to several dozen chapters. The time goes into preparation, not generation: splitting chapters, labelling speakers and making names consistent. A safe rhythm is to finish the first three chapters completely, confirm the narrator and the leads sound right, then format the rest in batches. To estimate how long a chapter will play, run its length through the script duration calculator.
In a novel, narration, dialogue and inner monologue are woven together. An audiobook pulls them apart: narration goes to the narrator at an even pace, dialogue goes to the characters and follows their emotions, and thoughts use the character's own voice, softer and closer. In a first-person novel, give the "I" narration to the protagonist's voice and it sounds like the hero telling their own story.

The easiest mistake to make
Don't paste novel-style attributions as they are. In a row like "She sneered: 'Get out.'", everything before the colon is read as a speaker name, so you get a character called "She sneered". Rewrite it as "Lin Wan (sneering): Get out." and both the speaker and the emotion come out right.
Each run is one chapter in its own project; the studio does not carry the previous chapter's cast over by itself, so your cast sheet is what keeps the book consistent. Use one spelling for every name, check each chapter's voices against the sheet after pasting, and update the sheet whenever you change something. More techniques are in keeping character voices consistent.
Turning someone else's novel into an audiobook and publishing it requires permission from the rights holder, meaning the author or a publisher licensed by them. Article 23 of China's Copyright Law protects an individual author's economic rights for their lifetime plus 50 years. After that the work is in the public domain and can be adapted freely.

One catch: a translation of a public-domain book may still be protected, because translators own the copyright in their translation, so recording an existing translation needs the translator's permission too. Audio platforms such as Ximalaya each have their own upload and copyright review rules; follow whatever they state at the time. In an August 2024 Xinhua report, an expert interviewed said that when users upload recordings of well-known works without valid proof of authorisation, platforms should refuse the upload. For the rules on using AI voices themselves, see can AI dubbing be used commercially.
The narrator will be in listeners' ears for many hours. Try three chapters first and only commit once it's still pleasant to hear.
Give frequent supporting characters voices far from the leads: if the heroine sounds young and bright, make her best friend more mature.
Shopkeepers and passers-by who appear once or twice can share a few "extra" voices, as long as none of them matches a main character.
Sect names, techniques and places are easy to misread. Fix their spelling while formatting and listen closely for them after generating.
For an English or other edition, generate and review one chapter first to settle name spellings and forms of address, then do the rest. See multi-character dubbing.
For your own listening, a phone or reading-app read-aloud is simplest. For a multi-voice audiobook you can publish, format each chapter as "Name: line", paste it into novel to audiobook software such as AI Drama Voice, generate chapter by chapter and export the MP3s.
In AI Drama Voice, cast analysis is free, so you can see the voice assignments first. Generation is charged by characters, and lines that fail are refunded automatically. See pricing for plans.
It depends on the number of chapters, the size of the cast and how fast you format the text. Generation runs per chapter; most of the time goes into splitting chapters, labelling speakers and listening back. Finish three chapters fully, then batch the rest.
Not if you keep a cast sheet: spell every name the same way and check each chapter's voices before generating. The studio doesn't carry one chapter's cast into the next on its own, so the sheet is what keeps a long book consistent.
Publishing someone else's novel needs the rights holder's permission. Public-domain books can be adapted freely, but an existing translation needs the translator's permission. Follow each platform's upload and copyright rules at the time.
Yes. Tick English when you generate: the lines are translated into English first, then voiced, and each character keeps the same voice in both editions. Each language exports its own MP3 and SRT.






Format a chapter, paste it, check the cast: a narrator and a voice for every character, MP3 and subtitles out, and the same cast sheet for the next chapter.
Billed by characters generated · Failed runs refunded automatically · Alipay, WeChat Pay and cards