Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -27,8 +27,8 @@ def translate(audio):
|
|
27 |
|
28 |
|
29 |
def synthesise(text):
|
|
|
30 |
max_length = processor.tokenizer.model_max_length
|
31 |
-
inputs = processor(text=text, max_length=max_length, truncation=True, return_tensors="pt")
|
32 |
speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
|
33 |
return speech.cpu()
|
34 |
|
@@ -40,10 +40,11 @@ def speech_to_speech_translation(audio):
|
|
40 |
return 16000, synthesised_speech
|
41 |
|
42 |
|
43 |
-
|
|
|
44 |
description = """
|
45 |
-
Demo for cascaded speech-to-speech translation (STST), mapping from source speech in
|
46 |
-
[
|
47 |
![Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")
|
48 |
"""
|
49 |
|
|
|
27 |
|
28 |
|
29 |
def synthesise(text):
|
30 |
+
inputs = processor(text=text, return_tensors="pt")
|
31 |
max_length = processor.tokenizer.model_max_length
|
|
|
32 |
speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
|
33 |
return speech.cpu()
|
34 |
|
|
|
40 |
return 16000, synthesised_speech
|
41 |
|
42 |
|
43 |
+
|
44 |
+
title = "English to Dutch Cascaded STST"
|
45 |
description = """
|
46 |
+
Demo for cascaded speech-to-speech translation (STST), mapping from source speech in English to target speech in Dutch. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and Microsoft's
|
47 |
+
and a finetuned SpeechT5[speecht5_tts_vox_nl](https://huggingface.co/sanchit-gandhi/speecht5_tts_vox_nl) model for text-to-speech:
|
48 |
![Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")
|
49 |
"""
|
50 |
|