File size: 2,799 Bytes
de1d53c e715e06 a1c5c13 e715e06 a1c5c13 e715e06 86df8f3 e715e06 a1c5c13 e715e06 956d8c0 e715e06 956d8c0 e715e06 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 |
import spaces
import gradio as gr
import librosa
import numpy as np
import torch
from transformers import pipeline
synthesiser = pipeline("text-to-speech", "techiaith/microsoft_speecht5_finetuned_bu_tts_cy_en")
speaker_embeddings = {
"GGP": "spkemb/speaker0.npy",
"BGP": "spkemb/speaker1.npy",
"BDP": "spkemb/speaker2.npy",
}
@spaces.GPU
def predict(text, speaker):
if len(text.strip()) == 0:
return (16000, np.zeros(0).astype(np.int16))
# limit input length
input_ids = inputs["input_ids"]
input_ids = input_ids[..., :model.config.max_text_positions]
if speaker == "Surprise Me!":
# load one of the provided speaker embeddings at random
idx = np.random.randint(len(speaker_embeddings))
key = list(speaker_embeddings.keys())[idx]
speaker_embedding = np.load(speaker_embeddings[key])
# randomly shuffle the elements
np.random.shuffle(speaker_embedding)
# randomly flip half the values
x = (np.random.rand(512) >= 0.5) * 1.0
x[x == 0] = -1.0
speaker_embedding *= x
#speaker_embedding = np.random.rand(512).astype(np.float32) * 0.3 - 0.15
else:
speaker_embedding = np.load(speaker_embeddings[speaker[:3]])
speaker_embedding = torch.tensor(speaker_embedding).unsqueeze(0)
speech = synthesiser(text, forward_params={"speaker_embeddings": speaker_embedding})
speech = (speech.numpy() * 32767).astype(np.int16)
return (16000, speech)
title = "Techiaith Finetune Microsoft/SpeechT5: Speech Synthesis"
description = """
Lleisiau TTS microsoft_speech_T5_finetune_bu_tts_cy_en
"""
examples = [
["Rhyfeddod neu ffenomenon optegol a meteorolegol yw enfys, pan fydd sbectrwm o olau yn ymddangos yn yr awyr pan fo'r haul yn disgleirio ar ddiferion o leithder yn atmosffer y ddaear.", "GGP (gwryw-gogledd-pro)"],
["Rhyfeddod neu ffenomenon optegol a meteorolegol yw enfys, pan fydd sbectrwm o olau yn ymddangos yn yr awyr pan fo'r haul yn disgleirio ar ddiferion o leithder yn atmosffer y ddaear.", "BGP (benyw-gogledd-pro)"],
["Rhyfeddod neu ffenomenon optegol a meteorolegol yw enfys, pan fydd sbectrwm o olau yn ymddangos yn yr awyr pan fo'r haul yn disgleirio ar ddiferion o leithder yn atmosffer y ddaear.", "BDP (benyw-de-pro)"],
]
gr.Interface(
fn=predict,
inputs=[
gr.Text(label="Input Text"),
gr.Radio(label="Speaker", choices=[
"GGP (gwryw-gogledd-pro)",
"BGP (benyw-gogledd-pro)",
"BDP (benyw-de-pro)",
"Surprise Me!"
],
value="GGP (gwryw-gogledd-pro)"),
],
outputs=[
gr.Audio(label="Generated Speech", type="numpy"),
],
title=title,
description=description,
examples=examples,
).launch()
|