Instructions to use yogenghodke/indic-f5-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use yogenghodke/indic-f5-mlx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir indic-f5-mlx yogenghodke/indic-f5-mlx
- F5-TTS
How to use yogenghodke/indic-f5-mlx with F5-TTS:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
File size: 1,122 Bytes
3285bd5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 | """Minimal IndicF5-MLX example.
Prereqs:
1. Apple-Silicon Mac
2. `huggingface-cli login` and accept the gate at https://huggingface.co/ai4bharat/IndicF5
3. A 24 kHz mono reference clip + its transcript
"""
import soundfile as sf
from indic_f5_mlx import load_indicf5, generate
REF_AUDIO = "reference_24k_mono.wav" # ~5-10s clean natural voice
REF_TEXT = "transcript of the reference clip in its language"
# Numbers spelled out as words (३७० -> तीनशे सत्तर), not digits.
TEXT = ("भारतीय संविधानाच्या अनुच्छेद तीनशे सत्तर नुसार, सव्वीस जानेवारी एकोणीसशे पन्नास रोजी "
"प्रजासत्ताक भारताची स्थापना झाली.")
model, _ = load_indicf5() # downloads + converts ai4bharat/IndicF5
audio = generate(model, ref_audio_path=REF_AUDIO, ref_text=REF_TEXT, text=TEXT, steps=16)
sf.write("out.wav", audio, 24000)
print(f"wrote out.wav ({len(audio)/24000:.1f}s)")
|