Update app.py

app.py

@@ -12,18 +12,19 @@ import json
 import soundfile as sf
 
 
+session_token = os.environ.get("SessionToken") 
+
 device = 0 if torch.cuda.is_available() else "cpu"
 
+# Intialise STT (Whisper)
 pipe = pipeline(
     task="automatic-speech-recognition",
-    model="openai/whisper-tiny.en",
+    model="openai/whisper-base.en",
     chunk_length_s=30,
     device=device,
-)
-
-session_token = os.environ.get("SessionToken")  
+) 
 
-# Intialize TTS (tacotron2) and Vocoder (HiFIGAN)
+# Intialise TTS (tacotron2) and Vocoder (HiFIGAN)
 tacotron2 = Tacotron2.from_hparams(source="speechbrain/tts-tacotron2-ljspeech", savedir="tmpdir_tts", overrides={"max_decoder_steps": 2000}, run_opts={"device":device})
 hifi_gan = HIFIGAN.from_hparams(source="speechbrain/tts-hifigan-ljspeech", savedir="tmpdir_vocoder")
 
@@ -46,7 +47,7 @@ def chat(input_audio, chat_history, reset_conversation):
     # text -> response (chatGPT)
     response = get_response_from_chatbot(message, reset_conversation)
 
-    # response -> speech (Tacotron 2)
+    # response -> speech (tacotron2)
     mel_output, mel_length, alignment = tacotron2.encode_text(response)
     wav = hifi_gan.decode_batch(mel_output)
     sf.write("out.wav", wav.squeeze().cpu().numpy(), 22050)
@@ -150,6 +151,7 @@ start_work= """async() => {
 
 with gr.Blocks(title="Talk to chatGPT") as demo:
     gr.Markdown("## Talk to chatGPT ##")
+    gr.HTML("<p> Demo uses <a href='https://huggingface.co/openai/whisper-base.en'>Whisper</a> to convert the input speech to transcribed text, <a href='https://chat.openai.com/chat'>chatGPT</a> to generate responses, and <a href='https://huggingface.co/speechbrain/tts-tacotron2-ljspeech'>tacotron2</a> to convert the response to output speech. </p>")
     gr.HTML("<p>You can duplicate this space and use your own session token: <a style='display:inline-block' href='https://huggingface.co/spaces/yizhangliu/chatGPT?duplicate=true'><img src='https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=10' alt='Duplicate Space'></a></p>")
     gr.HTML("<p> Instruction on how to get session token can be seen in video <a style='display:inline-block' href='https://www.youtube.com/watch?v=TdNSj_qgdFk'><font style='color:blue;weight:bold;'>here</font></a>. Add your session token by going to settings and add under secrets. </p>")
     with gr.Group(elem_id="page_1", visible=True) as page_1: