Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from gtts import gTTS
|
| 3 |
+
import tempfile
|
| 4 |
+
import difflib
|
| 5 |
+
import pandas as pd
|
| 6 |
+
from Levenshtein import distance as lev_distance
|
| 7 |
+
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
| 8 |
+
import torchaudio
|
| 9 |
+
|
| 10 |
+
# Load AI4Bharat Whisper model
|
| 11 |
+
processor = WhisperProcessor.from_pretrained("ai4bharat/indic-whisper-large-v2")
|
| 12 |
+
model = WhisperForConditionalGeneration.from_pretrained("ai4bharat/indic-whisper-large-v2").to("cpu") # use "cuda" if you have a GPU
|
| 13 |
+
|
| 14 |
+
# ... [play_text and helper functions as before] ...
|
| 15 |
+
|
| 16 |
+
def transcribe_audio(audio_path, original_text):
|
| 17 |
+
try:
|
| 18 |
+
# Load audio and preprocess
|
| 19 |
+
speech, rate = torchaudio.load(audio_path)
|
| 20 |
+
if rate != 16000:
|
| 21 |
+
resampler = torchaudio.transforms.Resample(orig_freq=rate, new_freq=16000)
|
| 22 |
+
speech = resampler(speech)
|
| 23 |
+
input_features = processor(speech.squeeze().numpy(), sampling_rate=16000, return_tensors="pt").input_features
|
| 24 |
+
|
| 25 |
+
# AI4Bharat Whisper transcription
|
| 26 |
+
predicted_ids = model.generate(input_features)
|
| 27 |
+
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0].strip()
|
| 28 |
+
|
| 29 |
+
# Error analysis and metrics (as before)
|
| 30 |
+
errors = compare_hindi_sentences(original_text, transcription)
|
| 31 |
+
df_errors = pd.DataFrame(errors, columns=["बिगड़ा हुआ शब्द", "संभावित सही शब्द", "गलती का प्रकार"])
|
| 32 |
+
|
| 33 |
+
duration = speech.shape[-1] / 16000
|
| 34 |
+
transcribed_words = transcription.strip().split()
|
| 35 |
+
speed = round(len(transcribed_words) / duration, 2) if duration > 0 else 0
|
| 36 |
+
accuracy = calculate_accuracy(original_text, transcription)
|
| 37 |
+
result_dict = {
|
| 38 |
+
"📝 Transcribed Text": transcription,
|
| 39 |
+
"⏱️ Speaking Speed (words/sec)": speed,
|
| 40 |
+
"✅ Reading Accuracy (%)": accuracy,
|
| 41 |
+
}
|
| 42 |
+
return result_dict, df_errors
|
| 43 |
+
except Exception as e:
|
| 44 |
+
return {"error": str(e)}, pd.DataFrame(columns=["बिगड़ा हुआ शब्द", "संभावित सही शब्द", "गलती का प्रकार"])
|