from transformers import WhisperProcessor, WhisperForConditionalGeneration
import soundfile as sf
import numpy as np
import os

Load model and processor

model_name = "openai/whisper-large" # Replace with your desired Whisper model
processor = WhisperProcessor.from_pretrained(model_name)
model = WhisperForConditionalGeneration.from_pretrained(model_name)

Load and process audio

audio_file = "converted_audio.wav" # Replace with your audio file path
audio, sampling_rate = sf.read(audio_file)

Split audio into 10-second chunks

chunk_duration = 10 # seconds
chunk_size = sampling_rate * chunk_duration
num_chunks = (len(audio) + chunk_size - 1) // chunk_size # Calculate number of chunks
chunks = [audio[i * chunk_size:(i + 1) * chunk_size] for i in range(num_chunks)]

Ensure output folder exists for temporary chunk files

temp_folder = "temp_audio_chunks"
os.makedirs(temp_folder, exist_ok=True)

Transcribe each chunk

transcriptions = []
for idx, chunk in enumerate(chunks):
temp_chunk_path = os.path.join(temp_folder, f"chunk_{idx + 1}.wav")
sf.write(temp_chunk_path, chunk, sampling_rate) # Save chunk to file

# Process the audio chunk
inputs = processor(chunk, sampling_rate=sampling_rate, return_tensors="pt", language="en")
input_features = inputs.input_features  # Use input_features instead of input_ids

# Generate the transcription using the input features (no need for attention_mask)
predicted_ids = model.generate(input_features)
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
transcriptions.append(transcription)
print(f"Transcribed chunk {idx + 1}/{num_chunks}: {transcription}")

Combine transcriptions

full_transcription = " ".join(transcriptions)

Save transcription to a text file

output_file = "transcription_output.txt"
with open(output_file, "w", encoding="utf-8") as file:
file.write(full_transcription)

Clean up temporary files

for chunk_file in os.listdir(temp_folder):
os.remove(os.path.join(temp_folder, chunk_file))
os.rmdir(temp_folder)

print(f"Full transcription saved to {output_file}")

this is the code that im using.