Beta
Explore
Marketplace
Neural Labs
Playground
Wallet
Docs
whisper-small-hindi-conversational – AI Model by YashikaBasapure | AlphaNeural AI
You can deploy this model and start earning money today!
YashikaBasapure
/
whisper-small-hindi-conversational
like
0
safetensors
whisper
hindi
conversational
asr
low-resource
fine-tuned
automatic-speech-recognition
hi
us
Views
No views yet
Model card
Files and Versions
Community
API
Deploy
Whisper Small — Fine-tuned on Conversational Hindi
Fine-tuned OpenAI Whisper-small on real conversational Hindi audio.
Results
Model
WER
Whisper-small baseline
178.08%
Fine-tuned (checkpoint-1200)
12.25%
After text normalisation
10.57%
Training Data
104 real speakers, long-form conversational recordings
4,929 segments after preprocessing (1s–30s filter)
90/10 train/test split
Novel Contributions
Lattice-based WER evaluation — handles Hindi filler variants (हाँ/हां/हम्म)
Hindi number normalisation pipeline
Roman-script English word detection for Hinglish audio
Training Config
Base model: openai/whisper-small
Steps: 2000 | Best checkpoint: 1200
Batch size: 16 | LR: 1e-5
How to Use
from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="YashikaBasapure/whisper-small-hindi-conversational") result = pipe("your_hindi_audio.wav") print(result["text"])
GitHub
https://github.com/Shishimanu9/Hindi-ASR
