Skip to main content
This solution is specially designed for novel reading platforms. By calling the core Large-Scale TTS API, it can convert a platform’s vast library of static text novels into high-quality audiobooks with rich emotions and multi-character performances in a one-click, large-scale process. This greatly enriches the content ecosystem and enhances user engagement.
Key Breakthroughs: Emotional consistency in long texts, multi-character voice generation, and second-level output for chapters with tens of thousands of words.
This solution aims to help partner platforms exponentially reduce audiobook production costs and significantly increase audio content coverage. It provides users with an immersive listening experience comparable to live human narration, thereby building a strong content moat.

​
Pain Points

We deeply understand the core challenges that novel platforms face in the audio content domain.

High Production Costs and Long Cycles

Traditional audiobook production requires hiring professional voice actors, recording studios, and post-production teams, leading to high costs (tens to hundreds of thousands of yuan) and long production cycles (weeks or even months) for a single book.

Flat, Monotonous Emotional Expression

Audio generated by traditional TTS (Text-to-Speech) technology lacks emotion and has a mechanical tone, failing to express the joy, anger, sorrow, and happiness of characters in a novel, resulting in a poor listening experience.

Lack of Context, Incoherent Narration

In long novels, standard TTS struggles to understand contextual connections, leading to disjointed character emotions and tones between chapters, which severely affects the story’s coherence and immersiveness.

Slow Content Updates, Low Coverage

Faced with a massive number of chapters updated daily on the platform, the speed of manual production is far from adequate. This results in the vast majority of works not having an audio version, thus missing out on a large audience of “listeners”.

Monotonous Voices, Difficult to Distinguish Characters

Platforms find it difficult to provide a sufficiently rich and personalized voice library to meet the stylistic needs of different novel genres (e.g., fantasy, romance, suspense). Distinguishing between multiple characters in a dialogue becomes particularly challenging.

​
Core Objectives

Ensure High Consistency of Emotion and Context

Achieve a deep understanding of long texts exceeding 10,000 characters to ensure that narration and character emotions remain coherent and natural throughout the entire chapter, aligning with the plot’s development.

Achieve Rich and Customizable Multi-Character Narration

Provide a large and continuously expanding “virtual voice actor” library that supports AI in automatically matching unique voices to different characters, while also allowing platforms for personalized customization.

Maximize the Production Efficiency of Audio Content

Compress the traditional weeks-long production cycle into minutes. Support single inputs of up to 35,000 characters, enabling “second-level” generation of entire novel chapters and making it possible to convert the entire site’s novel library into audio.

​
Solution

​
Step 1: Intelligent Full-Chapter Text Injection

Submit the novel chapter text to be converted (supports up to one million characters) in a single API call. The system will automatically perform text preprocessing, such as identifying chapter titles, narration, and dialogue.

​
Step 2: AI Director Intelligent Analysis

Our large model will act as an “AI Director”:
  • Contextual Understanding: Accurately parse the article’s intent, understanding character relationships and plot progression.
  • Character Recognition: Automatically identify different characters in the dialogue and match them with the most suitable voice from the voice library.
  • Emotion Analysis: Precisely identify the emotional tone of each sentence (e.g., excited, sad, tense, gentle) to provide “performance direction” for the subsequent speech synthesis.

​
Step 3: Dynamic Speech Synthesis and Delivery

Based on the analysis results, the AI performs the final speech synthesis:
  • Multi-Voice Fusion: Dynamically switch between different character voices, with steady narration and lively character portrayals.
  • Emotional Prosody: The generated speech is rich in variations of speed, pause, stress, and intonation, perfectly matching the text’s emotion.
  • Fast Delivery: Upon task completion, the API returns a URL for a high-quality MP3 audio file, ready for direct playback or distribution.

​
Business Value

Exponentially Boost Content Productivity

Reduce production costs by over 95% and increase production efficiency by 100 times. Quickly convert the platform’s entire novel assets into audio content, achieving a leap from “partial coverage” to “full coverage”.

Create an Excellent User Listening Experience

Provide a listening experience comparable to that meticulously crafted by a team of human professionals—multi-character, emotional, and consistent. Significantly increase average user listening time, completion rates, and paid conversion rates.

Build a Strong Content Differentiation Barrier

Rapidly launch a massive library of exclusive audiobooks to attract and retain the “listening” user base. Build a unique brand identity for the platform’s audiobooks by offering distinctive AI voices.

Ensure Data Privacy and Security Compliance

All text data is processed using strict encryption and desensitization techniques to ensure the security and compliance of the partner’s content assets and user data, providing peace of mind.

​
Core API Capabilities

This solution primarily relies on the following three API endpoints:

​
1. Create Audiobook Generation Task

POST https://api.minimax.io/v1/t2a_async_v2
  • Purpose: Create an asynchronous audiobook generation task. This is the core API call.
  • Key Parameters:
    • text The text content of the novel.
    • voice_setting Settings for speech synthesis, such as specifying multi-character mode, enabling/disabling emotion analysis, etc.
    • audio_setting Optional configuration to specify the preferred format for the generated audio.

​
2. Query Task Status

POST http://api.minimax.io/v1/query/t2a_async_query_v2
  • Purpose: Query the current status of a specified task (e.g., queued, processing, completed, failed).
  • Key Parameters:
    • task_id The unique ID returned when the task was created.

​
3. Get List of Available Voices

POST https://api.minimax.io/v1/get_voice
  • Purpose: Retrieve a list of all currently available AI voices and their characteristic tags (e.g., “Young Boy”, “Mature Female”, “Steady Uncle”, “Narrator”).
  • Use Case: To provide users with a voice selection feature.

​
Usage Example

The following is a Python code example for creating an audiobook generation task. 
import requests
import json

api_key = "Your API key here"

url = "https://api.minimax.io/v1/t2a_async_v2"
payload = {
    'role_voice':'{\
    "title": "audiobook_female_1",\
    "narration": "audiobook_male_1",\
    "Old Taoist Priest": "audiobook_male_2",\
    "Young Male": "gaodun_yanhan_kangfen"}',
    'interval_title': '1',
    'model':"speech-02",
}
files=[
  ('file',('shaosong1.json',open('C:\\Users\\minimax\\test_data\\shaosong1.json','rb'),'application/json'))
]
headers = {
  'Authorization': f'Bearer {api_key}',
}

response = requests.request("POST", url, headers=headers, data=payload, files=files)

print(response.text)