Kokoro Manim Voiceover

Professional library for generating high-quality voiceovers in Manim animations using the Kokoro-82M model.

Installation

From PyPI (Recommended)

pip install kokoro-manim-voiceover

Using uv (Alternative)

# For new projects
uv init my-project
cd my-project
uv add kokoro-manim-voiceover

# Or install directly
uv pip install kokoro-manim-voiceover

Note: uv is a modern, high-performance Python package manager that offers significantly faster installations and better dependency resolution than pip. It's fully compatible with existing Python workflows.

From Source

git clone https://github.com/xposed73/kokoro-manim-voiceover.git
cd kokoro-manim-voiceover
pip install -e .

Quick Start

from manim import *
from manim_voiceover import VoiceoverScene
from kokoro_mv import KokoroService

class MyAnimation(VoiceoverScene):
    def construct(self):
        self.set_speech_service(KokoroService(voice="af_sarah", lang="en-us"))
        
        with self.voiceover(text="Hello, this is my first voiceover!") as tracker:
            self.play(Write(Text("Hello, this is my first voiceover!")), run_time=tracker.duration)

Voices (Kokoro-82M)

This library uses the Kokoro-82M voice models. Below is a curated list of available voices grouped by language, adapted from the upstream model's VOICES.md.

Note: Voice quality varies by dataset size/quality. Some languages have fewer voices and may rely on fallback G2P. See the Kokoro-82M model card for details.

English (American)

af_heart, af_alloy, af_aoede, af_bella, af_jessica, af_kore,
af_nicole, af_nova, af_river, af_sarah, af_sky,
am_adam, am_echo, am_eric, am_fenrir, am_liam, am_michael,
am_onyx, am_puck, am_santa

English (British)

bf_alice, bf_emma, bf_isabella, bf_lily,
bm_daniel, bm_fable, bm_george, bm_lewis

Japanese

jf_alpha, jf_gongitsune, jf_nezumi, jf_tebukuro,
jm_kumo

Mandarin Chinese

zf_xiaobei, zf_xiaoni, zf_xiaoxiao, zf_xiaoyi,
zm_yunjian, zm_yunxi, zm_yunxia, zm_yunyang

Spanish

ef_dora, em_alex, em_santa

French

ff_siwis

Hindi

hf_alpha, hf_beta, hm_omega, hm_psi

Italian

if_sara, im_nicola

Brazilian Portuguese

pf_dora, pm_alex, pm_santa

Selecting voices and languages

• Set voice to one of the codes above.
• Set lang to match your target locale. Recommended codes for phonemizer compatibility:
  • English (US): en-us
  • English (GB): en-gb
  • Japanese: ja
  • Mandarin Chinese: cmn (instead of zh)
  • Spanish: es
  • French: fr-fr (instead of fr)
  • Hindi: hi
  • Italian: it
  • Brazilian Portuguese: pt-br

Example:

self.set_speech_service(KokoroService(voice="hf_alpha", lang="hi"))

Source: Kokoro-82M VOICES.md by hexgrad (Apache-2.0).

Video Samples

Preview recordings rendered from the included sample scenes:

Mandarin Chinese

https://github.com/user-attachments/assets/cc9cacd6-001e-43f8-a441-14ad20da6cfa

English (GB)

https://github.com/user-attachments/assets/12e01a68-35b6-435d-9cfe-5f414f16210a

English (US)

https://github.com/user-attachments/assets/d1b0de7a-42fb-4e46-a2c7-6ec0f2a37006

French

https://github.com/user-attachments/assets/f3c7eed1-b5c0-4423-a62a-19d56405d486

Hindi

https://github.com/user-attachments/assets/cc131788-03f7-4157-bb4e-db074f328d3e

Italian

https://github.com/user-attachments/assets/d2fd228e-0179-4008-9e5f-464b5cbfe053

Portuguese

https://github.com/user-attachments/assets/ed17f48f-e541-4b2a-b820-f0a13b175df8

Sample Scenes (code)

Code examples for each language are available in the samples/ folder. Each file sets an appropriate voice and lang and renders the same demo scene.

• samples/english_us.py → voice af_sarah, lang en-us
• samples/english_gb.py → voice bf_emma, lang en-gb
• samples/hindi.py → voice hf_alpha, lang hi
• samples/chinese_mandarin.py → voice zf_xiaobei, lang cmn
• samples/spanish.py → voice ef_dora, lang es
• samples/french.py → voice ff_siwis, lang fr-fr
• samples/italian.py → voice if_sara, lang it
• samples/portuguese_br.py → voice pf_dora, lang pt-br

Run any sample (replace the filename as needed):

# Option A: from repository root
manim -pql samples/english_us.py Example
# or with uv
uv run manim -pql samples/english_us.py Example

# Option B: cd into samples first
cd samples
manim -pql english_us.py Example

Notes:

• For Japanese and Mandarin, ensure eSpeak NG is installed and lang codes are set as above (e.g., cmn for Mandarin) to avoid phonemizer issues.
• If you encounter “phonemes too long” errors, split long narration into multiple with self.voiceover(...) blocks.
• The sample scenes use the manim-dsa library for data-structure visuals. Install it first if needed:

  pip install manim-dsa
  # OR
  uv add manim-dsa
  

Usage Examples

Basic Animation

class BasicExample(VoiceoverScene):
    def construct(self):
        self.set_speech_service(KokoroService(voice="af_sarah", lang="en-us"))
        
        circle = Circle()
        square = Square().shift(2 * RIGHT)
        
        with self.voiceover(text="This circle is drawn as I speak.") as tracker:
            self.play(Create(circle), run_time=tracker.duration)
        
        with self.voiceover(text="Now let's transform it into a square.") as tracker:
            self.play(Transform(circle, square), run_time=tracker.duration)

Custom Configuration

service = KokoroService(
    voice="af_sarah",    # Female voice
    speed=1.2,           # 20% faster
    lang="en-us",        # Language setting
    volume=1.2           # Volume
)

Requirements

• Python 3.11+
• Dependencies are automatically installed

Model Files

The library automatically downloads required model files (~220MB) on first use.

Development

Setting up development environment

Using uv (Recommended)

# Clone the repository
git clone https://github.com/xposed73/kokoro-manim-voiceover.git
cd kokoro-manim-voiceover

# Install in development mode with all dependencies
uv sync --dev

# Run tests
uv run pytest

# Format code
uv run black .
uv run isort .

# Type checking
uv run mypy .

Using pip

# Clone the repository
git clone https://github.com/xposed73/kokoro-manim-voiceover.git
cd kokoro-manim-voiceover

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install in development mode
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black .
isort .

Publishing to PyPI

To publish this package to PyPI:

1. Install build tools:

   pip install build twine
   

2. Build the package:

   python -m build
   

   This creates dist/ directory with wheel and source distribution files.

3. Upload to PyPI (TestPyPI first for testing):

   # Test on TestPyPI
   twine upload --repository-url https://test.pypi.org/legacy/ dist/*
   
   # Once verified, upload to production PyPI
   twine upload dist/*
   

4. Verify installation:

   pip install kokoro-manim-voiceover
   

Note: You'll need PyPI credentials (API token recommended). Create one at pypi.org/manage/account/token/

License

MIT License