Spaces:
Paused
Paused
Upload app.py
Browse files
app.py
CHANGED
|
@@ -1,58 +1,90 @@
|
|
| 1 |
import streamlit as st
|
| 2 |
-
import
|
| 3 |
-
import
|
| 4 |
-
from pathlib import Path
|
| 5 |
-
import base64
|
| 6 |
-
import io
|
| 7 |
-
import soundfile as sf
|
| 8 |
import numpy as np
|
|
|
|
|
|
|
|
|
|
| 9 |
from huggingface_hub import hf_hub_download
|
|
|
|
|
|
|
|
|
|
| 10 |
|
| 11 |
-
#
|
| 12 |
-
|
| 13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
-
#
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
config_path=config_file,
|
| 24 |
-
style_vec_path=style_file,
|
| 25 |
-
device="cpu" # クラウド環境ではCPUを使用
|
| 26 |
-
)
|
| 27 |
-
sr, wav = tts.infer(text=text, length=0.85)
|
| 28 |
|
| 29 |
-
#
|
| 30 |
-
|
| 31 |
-
sf.write(buffer, wav, sr, format='WAV')
|
| 32 |
-
audio_base64 = base64.b64encode(buffer.getvalue()).decode()
|
| 33 |
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
st.
|
| 37 |
-
|
| 38 |
-
#
|
| 39 |
-
st.
|
| 40 |
-
"https://hub.vroid.com/",
|
| 41 |
-
height=600,
|
| 42 |
-
scrolling=True
|
| 43 |
-
)
|
| 44 |
-
|
| 45 |
-
# チャットインターフェース
|
| 46 |
-
user_input = st.text_input("メッセージを入力してください:")
|
| 47 |
-
|
| 48 |
-
if user_input:
|
| 49 |
-
# AIvtuber.pyの機能を使用
|
| 50 |
-
resp = chat_session.send_message(user_input)
|
| 51 |
-
st.write("AI:", resp.text)
|
| 52 |
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import streamlit as st
|
| 2 |
+
import torch
|
| 3 |
+
from diffusers import StableBeluga2Pipeline
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
import numpy as np
|
| 5 |
+
import soundfile as sf
|
| 6 |
+
import io
|
| 7 |
+
import base64
|
| 8 |
from huggingface_hub import hf_hub_download
|
| 9 |
+
import asyncio
|
| 10 |
+
import time
|
| 11 |
+
from concurrent.futures import ThreadPoolExecutor
|
| 12 |
|
| 13 |
+
# モデルのダウンロードと初期化
|
| 14 |
+
@st.cache_resource
|
| 15 |
+
def load_model():
|
| 16 |
+
model_id = "stabilityai/stable-beluga-2"
|
| 17 |
+
pipe = StableBeluga2Pipeline.from_pretrained(
|
| 18 |
+
model_id,
|
| 19 |
+
torch_dtype=torch.float16,
|
| 20 |
+
use_safetensors=True,
|
| 21 |
+
variant="fp16"
|
| 22 |
+
)
|
| 23 |
+
if torch.cuda.is_available():
|
| 24 |
+
pipe = pipe.to("cuda")
|
| 25 |
+
return pipe
|
| 26 |
|
| 27 |
+
# 音声生成を非同期で実行する関数
|
| 28 |
+
def generate_audio_async(pipe, text, progress_bar):
|
| 29 |
+
try:
|
| 30 |
+
# 音声生成
|
| 31 |
+
audio = pipe(
|
| 32 |
+
text,
|
| 33 |
+
num_inference_steps=50,
|
| 34 |
+
guidance_scale=7.5
|
| 35 |
+
).audio[0]
|
| 36 |
+
|
| 37 |
+
# プログレスバーを更新
|
| 38 |
+
progress_bar.progress(1.0)
|
| 39 |
+
|
| 40 |
+
return audio
|
| 41 |
+
except Exception as e:
|
| 42 |
+
st.error(f"音声生成中にエラーが発生しました: {str(e)}")
|
| 43 |
+
return None
|
| 44 |
|
| 45 |
+
# メインアプリケーション
|
| 46 |
+
def main():
|
| 47 |
+
st.title("AI VTuber チャット")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
|
| 49 |
+
# モデルの読み込み
|
| 50 |
+
pipe = load_model()
|
|
|
|
|
|
|
| 51 |
|
| 52 |
+
# チャット履歴の初期化
|
| 53 |
+
if "messages" not in st.session_state:
|
| 54 |
+
st.session_state.messages = []
|
| 55 |
+
|
| 56 |
+
# チャット入力
|
| 57 |
+
user_input = st.chat_input("メッセージを入力してください")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
+
if user_input:
|
| 60 |
+
# ユーザーメッセージを表示
|
| 61 |
+
st.session_state.messages.append({"role": "user", "content": user_input})
|
| 62 |
+
|
| 63 |
+
# 音声生成の進捗バー
|
| 64 |
+
progress_bar = st.progress(0.0)
|
| 65 |
+
status_text = st.empty()
|
| 66 |
+
status_text.text("音声を生成中...")
|
| 67 |
+
|
| 68 |
+
# 音声生成を非同期で実行
|
| 69 |
+
with ThreadPoolExecutor() as executor:
|
| 70 |
+
future = executor.submit(generate_audio_async, pipe, user_input, progress_bar)
|
| 71 |
+
audio = future.result()
|
| 72 |
+
|
| 73 |
+
if audio is not None:
|
| 74 |
+
# 音声データを保存
|
| 75 |
+
audio_path = "generated_audio.wav"
|
| 76 |
+
sf.write(audio_path, audio.cpu().numpy(), 44100)
|
| 77 |
+
|
| 78 |
+
# 音声プレーヤーを表示
|
| 79 |
+
st.audio(audio_path)
|
| 80 |
+
|
| 81 |
+
# ステータステキストを更新
|
| 82 |
+
status_text.text("音声生成完了!")
|
| 83 |
+
|
| 84 |
+
# チャット履歴を表示
|
| 85 |
+
for message in st.session_state.messages:
|
| 86 |
+
with st.chat_message(message["role"]):
|
| 87 |
+
st.write(message["content"])
|
| 88 |
+
|
| 89 |
+
if __name__ == "__main__":
|
| 90 |
+
main()
|