Update app.py

app.py

@@ -1,51 +1,65 @@
-import gradio as gr
-import numpy as np
 import torch
-from datasets import load_dataset
+from transformers import pipeline
 
-from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+pipe = pipeline(
+    "automatic-speech-recognition", model="openai/whisper-base", device=device
+)
 
+# %%
 
-device = "cuda:0" if torch.cuda.is_available() else "cpu"
+def translate(audio):
+    outputs = pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "nl"})
+    return outputs["text"]
+
+# %%
 
-# load speech translation checkpoint
-pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device)
+from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
 
-# load text-to-speech checkpoint and speaker embeddings
 processor = SpeechT5Processor.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl")
 
-model = SpeechT5ForTextToSpeech.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl").to(device)
-vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
+model = SpeechT5ForTextToSpeech.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl")
+vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
 
-embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
-speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
+# %%
 
+model.to(device)
+vocoder.to(device)
 
-def translate(audio):
-    outputs = pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "nl"})
-    return outputs["text"]
+# %%
+
+from datasets import load_dataset, Audio
+
+embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
+speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
 
+# %%
 
 def synthesise(text):
     inputs = processor(text=text, return_tensors="pt")
-    speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
+    speech = model.generate_speech(
+        inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder
+    )
     return speech.cpu()
 
+# %%
+
+import numpy as np
+
+target_dtype = np.int16
+max_range = np.iinfo(target_dtype).max
+
 
 def speech_to_speech_translation(audio):
     translated_text = translate(audio)
     synthesised_speech = synthesise(translated_text)
-    synthesised_speech = (synthesised_speech.numpy() * 32767).astype(np.int16)
+    synthesised_speech = (synthesised_speech.numpy() * max_range).astype(np.int16)
     return 16000, synthesised_speech
 
 
-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")
-"""
+import gradio as gr
 
 demo = gr.Blocks()
 
@@ -53,17 +67,12 @@ mic_translate = gr.Interface(
     fn=speech_to_speech_translation,
     inputs=gr.Audio(source="microphone", type="filepath"),
     outputs=gr.Audio(label="Generated Speech", type="numpy"),
-    title=title,
-    description=description,
 )
 
 file_translate = gr.Interface(
     fn=speech_to_speech_translation,
     inputs=gr.Audio(source="upload", type="filepath"),
     outputs=gr.Audio(label="Generated Speech", type="numpy"),
-    examples=[["./example.wav"]],
-    title=title,
-    description=description,
 )
 
 with demo: