Update app.py

Add the code debugged on the local machine

app.py

@@ -5,13 +5,16 @@ from datasets import load_dataset
 
 from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
 
+from transliterate import translit
+
 
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
 
-# load speech translation checkpoint
+# загрузить контрольную точку модели транскибации и перевода речи
 asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device)
+translate_pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-en-ru")
 
-# load text-to-speech checkpoint and speaker embeddings
+# Загрузим контрольную точку преобразования текста в речь и эбеддинги дикторов
 processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
 
 model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
@@ -22,11 +25,15 @@ speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze
 
 
 def translate(audio):
-    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate"})
-    return outputs["text"]
+    transcription = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "english"})
+    translation = translate_pipe(transcription["text"])
+    result = translit(translation[0]['translation_text'], "ru", reversed=True)
+    print(result)
+    return result
 
 
 def synthesise(text):
+    print(text)
     inputs = processor(text=text, return_tensors="pt")
     speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
     return speech.cpu()
@@ -43,7 +50,6 @@ title = "Cascaded STST"
 description = """
 Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in English. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and Microsoft's
 [SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model for text-to-speech:
-
 ![Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")
 """
 
@@ -69,4 +75,4 @@ file_translate = gr.Interface(
 with demo:
     gr.TabbedInterface([mic_translate, file_translate], ["Microphone", "Audio File"])
 
-demo.launch()
+demo.launch()