Create app.py

app.py

@@ -0,0 +1,76 @@
+from faster_whisper import WhisperModel
+import pandas as pd
+import gradio as gr
+import psutil
+
+model = WhisperModel(whisper_model, device="cpu", compute_type="int8")
+
+def speech_to_text_translate(mic=None, file=None, lang=None):
+    if mic is not None:
+        audio = mic
+    elif file is not None:
+        audio = file
+    else:
+        raise gr.Error("You must either provide a mic recording or a file")
+    if lang is None:
+        raise gr.Error("Select a transcribe language")
+    
+    time_start = time.time()
+    segments, info = model.transcribe(audio, task='translate', beam_size=5)
+    print("Detected language '%s' with probability %f" % (info.language, info.language_probability))
+
+    #  Decode audio to Text
+    objects = [s._asdict() for s in segments]
+    time_end = time.time()
+    time_diff = time_end - time_start
+    memory = psutil.virtual_memory()
+    system_info = f"""
+    *Memory: {memory.total / (1024 * 1024 * 1024):.2f}GB, used: {memory.percent}%, available: {memory.available / (1024 * 1024 * 1024):.2f}GB.* 
+    *Processing time: {time_diff:.5} seconds.*
+    """
+    df_results = pd.DataFrame(objects)
+    df_results = df_results.drop(columns=['seek', 'tokens', 'avg_logprob'])
+    return df_results, system_info
+
+theme=gr.themes.Default().set(
+    color_accent="#e20074",
+
+    # Buttons
+    button_primary_text_color='white',
+    button_primary_text_color_hover='black',
+    button_primary_background_fill="#e20074",
+    button_primary_background_fill_hover='#c00063', # --telekom-color-primary-hovered
+    button_primary_border_color="#e20074",
+    button_primary_border_color_hover="#c00063",
+    stat_background_fill="#e20074",
+
+    # Dark Mode
+    button_primary_background_fill_dark="#e20074",
+    button_primary_background_fill_hover_dark='#c00063', # --telekom-color-primary-hovered
+    button_primary_border_color_dark="#e20074",
+    button_primary_border_color_hover_dark="#c00063",
+    stat_background_fill_dark="#e20074",
+)
+
+with gr.Blocks(title='Whisper Demo', theme=theme) as demo:
+    gr.Markdown('''
+        <div>
+        <h1 style='text-align: center'>Simple Whisper Demo</h1>
+        A simple Whisper demo using local CPU Inference of the largest-v2 Model
+        </div>
+    ''')
+    audio_in = gr.Audio(label="Record", source='microphone', type="filepath")
+    file_in = gr.Audio(label="Upload", source='upload', type="filepath")
+    drop_down = gr.Dropdown(["de", "en", "es", "fr", "ru"], value="en")
+    transcribe_btn = gr.Button("Transcribe audio", variant="primary")
+    translate_btn = gr.Button("Translate audio")
+    trans_df = gr.DataFrame(label="Transcription dataframe", row_count=(0, "dynamic"), max_rows = 10, wrap=True, overflow_row_behaviour='paginate')
+    sys_info = gr.Markdown(f"*Memory: {memory.total / (1024 * 1024 * 1024):.2f}GB, used: {memory.percent}%, available: {memory.available / (1024 * 1024 * 1024):.2f}GB*")
+    transcribe_btn.click(speech_to_text_simple, 
+                                 [audio_in, file_in], 
+                                 [trans_df, sys_info]
+                                )
+    translate_btn.click(speech_to_text_translate, 
+                            [audio_in, file_in, drop_down], 
+                            [trans_df, sys_info]
+