IndicF5: High-Quality Text-to-Speech for Indian Languages

We release IndicF5, a near-human polyglot Text-to-Speech (TTS) model trained on 1417 hours of high-quality speech from Rasa, IndicTTS, LIMMITS, and IndicVoices-R.

IndicF5 supports 11 Indian languages:
Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, Telugu.

NOTE - All rights goes to AI4Bharact, this repository is meant to improve the documentation and ease of installation.

🚀 Installation

1. Install the ffmpeg library

sudo apt install ffmpeg

For CPU, download the ffmpeg and add the path to the bin directory in the system variable path. Follow the instructions below to install ffmpeg on CPU

https://www.wikihow.com/Install-FFmpeg-on-Windows

- Download the ffmpeg for windows from https://github.com/GyanD/codexffmpeg/releases/download/7.1.1/ffmpeg-7.1.1-full_build-shared.7z. Please note that torchcodec works with ffmpeg version <=8.0
- Extract the zip file and copy the path to bin file within the folder.
- Add the path to ffmpeg bin files in the System Variables.

2. Create the environment and install dependencies:

   • CPU

   uv venv python=3.10
   source .venv/bin/activate
   uv pip install git+https://github.com/ai4bharat/IndicF5.git --torch-backend=cpu
   uv pip install torchcodec transformers==4.50.0 --torch-backend=cpu
   

   • GPU

   uv venv python=3.10
   source .venv/bin/activate
   uv pip install git+https://github.com/ai4bharat/IndicF5.git
   uv pip install torchcodec transformers==4.50.0 
   

🎙 Usage

To generate speech, you need to provide three inputs:

1. Text to synthesize – The content you want the model to speak.
2. A reference prompt audio – An example speech clip that guides the model’s prosody and speaker characteristics.
3. Text spoken in the reference prompt audio – The transcript of the reference prompt audio.

• Download the reference audio file:

wget https://huggingface.co/Aditya02/IndicF5/resolve/main/prompts/PAN_F_HAPPY_00001.wav

• Code

from transformers import AutoModel
import numpy as np
import soundfile as sf

# Load IndicF5 from Hugging Face
repo_id = "Aditya02/IndicF5"
model = AutoModel.from_pretrained(repo_id, trust_remote_code=True)

# Generate speech
audio = model(
    "नमस्ते! संगीत की तरह जीवन भी खूबसूरत होता है, बस इसे सही ताल में जीना आना चाहिए.",
    ref_audio_path="PAN_F_HAPPY_00001.wav",
    ref_text="ਭਹੰਪੀ ਵਿੱਚ ਸਮਾਰਕਾਂ ਦੇ ਭਵਨ ਨਿਰਮਾਣ ਕਲਾ ਦੇ ਵੇਰਵੇ ਗੁੰਝਲਦਾਰ ਅਤੇ ਹੈਰਾਨ ਕਰਨ ਵਾਲੇ ਹਨ, ਜੋ ਮੈਨੂੰ ਖੁਸ਼ ਕਰਦੇ  ਹਨ।"
)

# Normalize and save output
if audio.dtype == np.int16:
    audio = audio.astype(np.float32) / 32768.0
sf.write("namaste.wav", np.array(audio, dtype=np.float32), samplerate=24000)
print("Audio saved succesfully.")

You can find example prompt audios used here.

Terms of Use

By using this model, you agree to only clone voices for which you have explicit permission. Unauthorized voice cloning is strictly prohibited. Any misuse of this model is the responsibility of the user.

References

We would like to extend our gratitude to the authors of F5-TTS for their invaluable contributions and inspiration to this work. Their efforts have played a crucial role in advancing the field of text-to-speech synthesis.

📖 Citation

If you use IndicF5 in your research or projects, please consider citing it:

🔹 BibTeX

@misc{AI4Bharat_IndicF5_2025,
  author       = {Praveen S V and Srija Anand and Soma Siddhartha and Mitesh M. Khapra},
  title        = {IndicF5: High-Quality Text-to-Speech for Indian Languages},
  year         = {2025},
  url          = {https://github.com/AI4Bharat/IndicF5},
}