Upload 3 files

app.py

@@ -0,0 +1,34 @@
+import gradio as gr
+from ts_utilites import transcribe_speech, transcribe_long_form, text_to_speech
+
+# Gradio App
+app = gr.TabbedInterface(
+    [
+        # Transcribe Speech
+        gr.Interface(
+            fn=transcribe_speech,
+            inputs=gr.Audio(type="filepath"),
+            outputs=gr.Textbox(label="Transcription", lines=5),
+            title="Transcribe Speech",
+            allow_flagging="never",
+        ),
+        # Long-Form Transcription
+        gr.Interface(
+            fn=transcribe_long_form,
+            inputs=gr.Audio(type="filepath"),
+            outputs=gr.Textbox(label="Transcription", lines=10),
+            title="Long-Form Transcription",
+            allow_flagging="never",
+        ),
+        # Text-to-Speech
+        gr.Interface(
+            fn=text_to_speech,
+            inputs=gr.Textbox(label="Enter Text", placeholder="Type your text here...", lines=5),
+            outputs=gr.Audio(label="Generated Speech"),
+            title="Text-to-Speech",
+            allow_flagging="never",
+        )
+    ],
+    ["Transcribe Speech", "Long-Form Transcription", "Text-to-Speech"]
+)
+app.launch()

requirements.txt

@@ -0,0 +1,6 @@
+gradio
+transformers
+torch
+librosa
+soundfile
+phonemizer

ts_utilities.py

@@ -0,0 +1,51 @@
+import numpy as np
+import librosa
+import soundfile as sf
+from datasets import load_dataset
+from transformers import pipeline
+
+# Initialize pipelines for speech recognition and tts models
+asr = pipeline(task="automatic-speech-recognition", model="distil-whisper/distil-small.en")
+narrator = pipeline("text-to-speech", model="kakao-enterprise/vits-ljs")
+
+# Speech-to-Text Function
+def transcribe_speech(filepath):
+    if filepath is None:
+        return "No audio found. Please retry."
+    output = asr(filepath)
+    return output["text"]
+
+# Long-Form Audio Transcription
+def transcribe_long_form(filepath):
+    if filepath is None:
+        return "No audio found. Please retry."
+    # Load and preprocess audio
+    audio, sampling_rate = sf.read(filepath)
+    audio_transposed = np.transpose(audio)
+    audio_mono = librosa.to_mono(audio_transposed)
+    audio_16KHz = librosa.resample(audio_mono, orig_sr=sampling_rate, target_sr=16000)
+    # Transcribe using ASR pipeline
+    chunks = asr(audio_16KHz, chunk_length_s=30, batch_size=4, return_timestamps=True)["chunks"]
+    # Combine all transcriptions
+    return "\n".join([chunk["text"] for chunk in chunks])
+
+# Text-to-Speech Function
+def text_to_speech(text):
+    if not text.strip():
+        return "No text provided. Please enter text to synthesize."
+    narrated_text = narrator(text)
+    audio_array = narrated_text["audio"][0].flatten()  # Flatten the 2D array to 1D
+    sampling_rate = narrated_text["sampling_rate"]  # Get sampling rate
+    return sampling_rate, audio_array
+
+# Sample Dataset Access Function
+def get_dataset_sample(idx):
+    dataset = load_dataset("librispeech_asr", split="train.clean.100", streaming=True, trust_remote_code=True)
+    # example = next(iter(dataset))  
+    dataset_head = list(dataset.take(5)) 
+    sample = dataset_head[idx]
+    audio_array = sample["audio"]["array"]
+    sampling_rate = sample["audio"]["sampling_rate"]
+    transcription = sample["text"]
+    return (audio_array, sampling_rate), transcription
+