NavaiSTT-1v Medium: Uzbek language Fine-tuned Whisper medium Speech-to-Text model

NavaiSTT-1v Medium

Classic Whisper medium model fine-tuned for Uzbek language. The dataset included ~700 hours of diverse audio: publicly available podcasts, Tashkent dialect podcasts, audiobooks, and Common Voice 17. Data quality was mixed with 60% human transcribed and 40% pseudo-transcribed using Gemini 2.5 Pro.

Special attention was given to Tashkent dialect audio materials, resulting in strong performance on this dialect. Future versions will include other regional dialects to improve overall coverage.

Original Model: Whisper Medium
Parameters: 769 M
Wer: ~13%
Cer: ~3.5%

Tensorbaord metrics for: NavaiSTT-1v Medium

This is only the first version of this model. Based on how helpful it proves for users, I will begin work on v2. I already have a clear vision for how to improve the model further.

Performance Comparison

Listen to the sample below and compare NavaiSTT-1v Medium transcription accuracy with other leading services:

Thoughts

Let's be honest - most startups don't need open source AI models unless they are facing specific legal restrictions. Gemini 1.5, 2, 2.5 Pro delivers remarkably human-like performance, especially for transcription where it captures nuanced speech with accuracy that rivals human ears. When you hear someone speaking, this model understands exactly as you would.

Commercial solutions like Gemini are production-ready and surprisingly affordable for most business applications. The transcription quality is so refined that it processes audio similar to how human listeners perceive conversations - catching subtle inflections and handling background noise gracefully.

However, when handling sensitive personal data that cannot be shared with foreign third parties, such as private audio recordings containing confidential information, regionally-hosted services or self-hosted open source alternatives (like those based in Uzbekistan) provide necessary compliance without compromising too much on quality.

Implementation Example

Integrating NavaiSTT into your application is straightforward. Here's a Python code example:

import torch
import torchaudio
import numpy as np
from transformers import WhisperProcessor, WhisperForConditionalGeneration

def transcribe_audio(audio_path, model_path):
    """Simple function to transcribe audio with a fine-tuned Whisper model"""
    # Load model and processor
    processor = WhisperProcessor.from_pretrained(model_path)
    model = WhisperForConditionalGeneration.from_pretrained(model_path)

    # Move to GPU if available
    device = "cuda" if torch.cuda.is_available() else "cpu"
    model = model.to(device)

    # Load and preprocess audio
    waveform, sample_rate = torchaudio.load(audio_path)
    if sample_rate != 16000:
        waveform = torchaudio.functional.resample(waveform, sample_rate, 16000)

    # Convert to mono if needed
    if waveform.shape[0] > 1:
        waveform = waveform.mean(dim=0, keepdim=True)

    # Process audio
    input_features = processor(
        waveform.squeeze().numpy(),
        sampling_rate=16000,
        return_tensors="pt"
    ).input_features.to(device)

    # Generate transcription
    with torch.no_grad():
        predicted_ids = model.generate(input_features)

    # Decode
    transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
    return transcription

# Example usage
if __name__ == "__main__":
    MODEL_PATH = "islomov/navaistt_v1_medium"
    audio_file = "some_audio.wav"

    text = transcribe_audio(audio_file, MODEL_PATH)
    print(f"Transcription: {text}")