Update app.py

app.py

@@ -139,7 +139,7 @@ pipe = pipeline(
     chunk_length_s=30,
     device=device,
 )
-
+os.makedirs('output', exist_ok=True)
 pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language=lang, task="transcribe")
 
 embedding_model = PretrainedSpeakerEmbedding( 
@@ -286,8 +286,10 @@ def speech_to_text(video_file_path, selected_source_lang, whisper_model, num_spe
         *Processing time: {time_diff:.5} seconds.*
         *GPU Utilization: {gpu_utilization}%, GPU Memory: {gpu_memory}MiB.*
         """
-
-        return pd.DataFrame(objects), system_info
+        save_path = "output/transcript_result.csv"
+        df_results = pd.DataFrame(objects)
+        df_results.to_csv(save_path)
+        return df_results, system_info, save_path
     
     except Exception as e:
         raise RuntimeError("Error Running inference with local model", e)
@@ -303,6 +305,7 @@ selected_source_lang = gr.Dropdown(choices=source_language_list, type="value", v
 selected_whisper_model = gr.Dropdown(choices=whisper_models, type="value", value="base", label="Selected Whisper model", interactive=True)
 number_speakers = gr.Number(precision=0, value=2, label="Selected number of speakers", interactive=True)
 system_info = gr.Markdown(f"*Memory: {memory.total / (1024 * 1024 * 1024):.2f}GB, used: {memory.percent}%, available: {memory.available / (1024 * 1024 * 1024):.2f}GB*")
+download_transcript = gr.File(label="Download transcript")
 transcription_df = gr.DataFrame(value=df_init,label="Transcription dataframe", row_count=(0, "dynamic"), max_rows = 10, wrap=True, overflow_row_behaviour='paginate')
 title = "Whisper speaker diarization"
 demo = gr.Blocks(title=title)
@@ -358,8 +361,9 @@ with demo:
                 selected_source_lang.render()
                 selected_whisper_model.render()
                 number_speakers.render()
+                download_transcript.render()
                 transcribe_btn = gr.Button("Transcribe audio and diarization")
-                transcribe_btn.click(speech_to_text, [video_in, selected_source_lang, selected_whisper_model, number_speakers], [transcription_df, system_info])
+                transcribe_btn.click(speech_to_text, [video_in, selected_source_lang, selected_whisper_model, number_speakers], [transcription_df, system_info, download_transcript])
 
                 
         with gr.Row():