|
from io import StringIO
|
|
import time
|
|
import os
|
|
import logging
|
|
|
|
import gradio as gr
|
|
import pandas as pd
|
|
from pypinyin import lazy_pinyin
|
|
from gradio_i18n import , Translate
|
|
|
|
from api import generate_api
|
|
|
|
|
|
trans_file = os.path.join(os.path.dirname(__file__),"i18n", "translations.json")
|
|
|
|
|
|
logging.getLogger('aiohttp').setLevel(logging.WARNING)
|
|
logging.getLogger("gradio").setLevel(logging.WARNING)
|
|
|
|
|
|
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
|
|
|
|
|
terms = r"""
|
|
## 免责声明
|
|
|
|
本网站提供的语音合成服务（以下简称“服务”）旨在供个人使用和娱乐目的。在使用本服务前，请用户仔细阅读并充分理解以下条款：
|
|
|
|
1. **角色版权**：本网站可能使用的角色形象涉及第三方知识产权。本网站不拥有这些角色的版权。用户在使用服务时应尊重相关角色的知识产权，并确保其行为不侵犯任何第三方的知识产权。
|
|
|
|
2. **用户生成内容（UGC）**：用户通过本平台生成的语音内容（以下简称“UGC”）由用户自行负责，与本平台无关。本平台无法控制或审核用户生成的具体内容，且不对UGC的准确性、完整性或合法性承担任何责任。
|
|
|
|
3. **使用限制**：本服务生成的语音及其UGC仅限于个人使用，不得用于任何商业目的。未经本平台事先书面同意，禁止将生成内容用于任何商业活动。
|
|
|
|
4. **法律责任**：用户使用本服务所产生的任何法律责任由用户自行承担，与本平台无关。如因用户使用服务或其UGC导致的任何纠纷或损失，本平台不承担任何责任。
|
|
|
|
5. **版权声明**：用户应尊重原创，不得使用本服务生成侵犯他人著作权的内容。如发现用户生成内容侵犯他人版权，本平台有权立即停止对其提供服务，并保留追究法律责任的权利。
|
|
|
|
6. **内容监管**：尽管本平台无法控制UGC，但一旦发现违反本免责声明或法律法规的内容，本平台将采取必要措施，包括但不限于删除违规内容，并配合有关部门进行调查。
|
|
|
|
7. **注明要求**：用户应在生成内容的显著位置，如可能的话，注明“此内容由RubiiTTS生成”或类似的说明。用户应确保注明行为符合本条款的要求。
|
|
|
|
用户使用本网站即表示同意以上免责声明。如有疑问，请联系我们contact@yomio.ai。
|
|
|
|
**最终解释权归本网站所有。**
|
|
|
|
"""
|
|
|
|
terms_js = r"alert('本网站提供的语音合成服务仅供个人使用和娱乐目的。请注意以下几点:\n1. 角色版权:本网站使用的角色形象可能涉及第三方知识产权,我们不拥有这些角色的版权。\n2. 生成内容:用户通过本平台生成的语音内容由用户自行负责,与本平台无关。我们无法控制或审核用户生成的内容。\n3. 使用限制:生成的语音仅限个人使用,不得用于任何商业目的。\n4. 法律责任:用户使用本服务所产生的任何法律责任由用户自行承担,与本平台无关。\n5. 版权声明:请尊重原创,不要使用本服务生成侵犯他人著作权的内容。\n使用本网站即表示您同意以上免责声明。如有疑问,请联系我们。')"
|
|
|
|
def load_characters_csv(lang):
|
|
name = f"characters_{lang}"
|
|
return pd.read_csv(StringIO(os.getenv(name)))
|
|
|
|
def update_all_characters(lang, current_all_characters):
|
|
new_characters = load_characters_csv(lang)
|
|
initial_characters = get_characters(kind="原神", all_characters=new_characters)
|
|
return new_characters, initial_characters, gr.Gallery(value=[[char['头像'], char['名称']] for char in initial_characters],
|
|
show_label=False, elem_id="character_gallery", columns=[11],
|
|
object_fit="contain", height="auto", interactive=False,
|
|
allow_preview=False, selected_index=None)
|
|
|
|
def get_characters(query=None, page=1, per_page=400, kind="原神", lang="zh", all_characters=None):
|
|
|
|
filtered_characters = all_characters[all_characters["类别"] == kind]
|
|
|
|
if query:
|
|
|
|
filtered_characters = filtered_characters[
|
|
filtered_characters['名称'].str.contains(query, case=False)
|
|
]
|
|
if filtered_characters.empty and lang == 'zh':
|
|
filtered_characters = all_characters[all_characters["类别"] == kind]
|
|
filtered_characters = filtered_characters[
|
|
filtered_characters['名称'].apply(lambda x: ''.join(lazy_pinyin(x))).str.contains(query, case=False)
|
|
]
|
|
|
|
|
|
unique_characters = filtered_characters.groupby('名称').first().reset_index().sort_values(by='id')
|
|
|
|
|
|
start_index = (page - 1) * per_page
|
|
end_index = start_index + per_page
|
|
|
|
return unique_characters.iloc[start_index:end_index].to_dict('records')
|
|
|
|
async def generate(selected_character = None, selected_characters = [], text = "", lang="zh"):
|
|
|
|
|
|
if selected_character:
|
|
characters = [selected_character] + selected_characters
|
|
else:
|
|
characters = selected_characters
|
|
if not selected_character and not selected_characters:
|
|
if lang == "zh":
|
|
raise gr.Error("请先选择一个角色")
|
|
elif lang == "en":
|
|
raise gr.Error("Please select a character first")
|
|
elif lang == "ja":
|
|
raise gr.Error("まず、キャラクターを選択してください")
|
|
elif lang == "ko":
|
|
raise gr.Error("먼저 캐릭터를 선택하세요")
|
|
voice_ids = [char.get("voice_id") for char in characters if char.get("voice_id")]
|
|
|
|
if not voice_ids:
|
|
raise gr.Error("所选角色没有关联的 voice_id")
|
|
|
|
start_time = time.time()
|
|
|
|
if voice_ids == "1":
|
|
if lang == "zh":
|
|
raise gr.Error("该角色暂未创建语音")
|
|
elif lang == "en":
|
|
raise gr.Error("The character has not been created yet")
|
|
elif lang == "ja":
|
|
raise gr.Error("そのキャラクターの音声はまだ作成されていません")
|
|
elif lang == "ko":
|
|
raise gr.Error("해당 캐릭터의 음성이 아직 생성되지 않았습니다")
|
|
|
|
if text == "":
|
|
if lang == "zh":
|
|
raise gr.Error("请输入需要合成的文本")
|
|
elif lang == "en":
|
|
raise gr.Error("Please enter the text to be synthesized")
|
|
elif lang == "ja":
|
|
raise gr.Error("合成するテキストを入力してください")
|
|
elif lang == "ko":
|
|
raise gr.Error("합성할 텍스트를 입력하세요")
|
|
|
|
if (lang == "en" and len(text.split()) > 200) or len(text) > 512:
|
|
if lang == "zh":
|
|
raise gr.Error("长度请控制在512个字符以内")
|
|
elif lang == "en":
|
|
raise gr.Error("The text length exceeds 200 words")
|
|
elif lang == "ja":
|
|
raise gr.Error("テキストの長さが512文字を超えています")
|
|
elif lang == "ko":
|
|
raise gr.Error("텍스트 길이가 512자를 초과합니다")
|
|
|
|
|
|
audio = await generate_api(voice_ids, text)
|
|
end_time = time.time()
|
|
if lang == "zh":
|
|
cost_time = f"合成共花费{end_time - start_time:.2f}秒"
|
|
elif lang == "en":
|
|
cost_time = f"Total time spent synthesizing: {end_time - start_time:.2f} seconds"
|
|
elif lang == "ja":
|
|
cost_time = f"合成にかかった時間: {end_time - start_time:.2f}秒"
|
|
elif lang == "ko":
|
|
cost_time = f"합성에 소요된 시간: {end_time - start_time:.2f}초"
|
|
if isinstance(audio, str):
|
|
print(audio)
|
|
raise gr.Error(audio)
|
|
else:
|
|
return audio, cost_time
|
|
|
|
def get_character_emotions(character, all_characters):
|
|
|
|
character_records = all_characters[all_characters['名称'] == character['名称']]
|
|
|
|
|
|
emotions = character_records['情绪'].unique().tolist()
|
|
|
|
|
|
return emotions if emotions else ["默认情绪"]
|
|
|
|
def update_character_info(character_name, emotion, current_character, all_characters):
|
|
character_info = None
|
|
if character_name and emotion:
|
|
character_info = all_characters[(all_characters['名称'] == character_name) & (all_characters['情绪'] == emotion)]
|
|
if character_name == "":
|
|
return None
|
|
character_info = character_info.iloc[0].to_dict()
|
|
return character_info, all_characters
|
|
|
|
def add_new_voice(current_character, selected_characters, kind, lang, all_characters):
|
|
if not current_character:
|
|
if lang == "zh":
|
|
raise gr.Error("请先选择一个角色")
|
|
elif lang == "en":
|
|
raise gr.Error("Please select a character first")
|
|
elif lang == "ja":
|
|
raise gr.Error("まず、キャラクターを選択してください")
|
|
elif lang == "ko":
|
|
raise gr.Error("먼저 캐릭터를 선택하세요")
|
|
|
|
if len(selected_characters) >= 5:
|
|
raise gr.Error("已达到最大选择数（5个）")
|
|
|
|
|
|
existing_char = next((char for char in selected_characters if char['名称'] == current_character['名称']), None)
|
|
if existing_char:
|
|
|
|
if existing_char['情绪'] != current_character['情绪']:
|
|
existing_char['情绪'] = current_character['情绪']
|
|
else:
|
|
selected_characters.insert(0, current_character)
|
|
|
|
updated_characters = get_characters(kind=kind, lang=lang, all_characters=all_characters)
|
|
|
|
updated_gallery = gr.Gallery(value=[[char['头像'], char['名称']] for char in updated_characters],
|
|
show_label=False, elem_id="character_gallery", columns=[11],
|
|
object_fit="contain", height="auto", interactive=False,
|
|
allow_preview=False, selected_index=None)
|
|
|
|
return (None, gr.update(value=""), gr.update(choices=[]), selected_characters,
|
|
updated_characters, updated_gallery, gr.update(visible=True), all_characters)
|
|
|
|
def update_selected_chars_display(selected_characters):
|
|
updates = []
|
|
for i, (name, emotion, _, row) in enumerate(selected_chars_rows):
|
|
if i < len(selected_characters):
|
|
char = selected_characters[i]
|
|
updates.extend([
|
|
gr.update(value=char['名称'], visible=True),
|
|
gr.update(value=char['情绪'], visible=True),
|
|
gr.update(visible=True),
|
|
gr.update(visible=True)
|
|
])
|
|
else:
|
|
updates.extend([
|
|
gr.update(value="", visible=False),
|
|
gr.update(value="", visible=False),
|
|
gr.update(visible=False),
|
|
gr.update(visible=False)
|
|
])
|
|
return updates
|
|
|
|
def remove_character(index, selected_characters):
|
|
if 0 <= index < len(selected_characters):
|
|
del selected_characters[index]
|
|
return selected_characters, gr.update(visible=True)
|
|
|
|
def update_gallery(kind, query, all_characters):
|
|
updated_characters = get_characters(kind=kind, query=query, lang=lang, all_characters=all_characters)
|
|
return updated_characters, [[char['头像'], char['名称']] for char in updated_characters], all_characters
|
|
|
|
def on_select(evt: gr.SelectData, characters, selected_characters, all_characters):
|
|
|
|
if len(selected_characters) == 0:
|
|
selected_characters = []
|
|
|
|
selected = characters[evt.index]
|
|
emotions = get_character_emotions(selected, all_characters)
|
|
default_emotion = emotions[0] if emotions else ""
|
|
|
|
character_dict = selected.copy()
|
|
character_dict['情绪'] = default_emotion
|
|
|
|
return selected["名称"], gr.Dropdown(choices=emotions, value=default_emotion), character_dict, selected_characters
|
|
|
|
with gr.Blocks(title="Rubii TTS", theme=gr.themes.Soft()) as demo:
|
|
lang = gr.Radio(choices=[("中文", "zh"), ("English", "en"), ("日本語", "ja"), ("한국인", "ko")], label=("Language"), value="zh", scale=1)
|
|
all_characters_state = gr.State(load_characters_csv("zh"))
|
|
|
|
|
|
gr.Markdown(
|
|
value=("""## 🎉 欢迎使用Rubii语音合成系统 🎉
|
|
|
|
#### [🗣️ 不想只是听到角色的声音，还想与他们进行互动交流吗？快点击我来体验与这些角色的生动对话吧！(中国大陆暂不可用) 🌟](https://rubii.ai)
|
|
|
|
📝 使用说明:
|
|
1. 选择角色类别 🎭
|
|
2. 从图库中选择一个或多个角色(最多5个) 👥。当选择多个角色时，系统会自动进行声线融合(以第一个角色为主音色，其他角色为辅助音色)，您可以尝试不同的组合来获得独特的声音效果。
|
|
3. 选择角色的情绪 😊😢😠
|
|
4. 输入要合成的文本 ✍️
|
|
5. 点击"合成语音"按钮 🔊
|
|
"""
|
|
))
|
|
with gr.Group():
|
|
initial_characters = get_characters(kind="原神", lang="zh", all_characters=all_characters_state.value)
|
|
characters = gr.State(initial_characters)
|
|
selected_characters = gr.State([])
|
|
current_character = gr.State(None)
|
|
|
|
with gr.Blocks():
|
|
with gr.Row():
|
|
|
|
choices = ["原神", "崩坏星穹铁道", "鸣潮"]
|
|
kind = gr.Dropdown(choices=[((name), name) for name in choices], value="原神", label=("选择角色类别"))
|
|
query = gr.Textbox(label=("搜索角色"), value="", lines=1, max_lines=1, interactive=True)
|
|
with gr.Blocks():
|
|
gallery = gr.Gallery(
|
|
value=[[char['头像'], char['名称']] for char in characters.value],
|
|
show_label=False,
|
|
elem_id="character_gallery",
|
|
columns=[11],
|
|
object_fit="contain",
|
|
height="auto",
|
|
interactive=False,
|
|
allow_preview=False,
|
|
selected_index=None
|
|
)
|
|
with gr.Row():
|
|
character_name = gr.Textbox(label=("当前选择的角色"), interactive=False, max_lines=1)
|
|
info_type = gr.Dropdown(choices=[], label=("选择情绪"))
|
|
with gr.Row():
|
|
add_voice_button = gr.Button(("添加新的声音"), variant="primary")
|
|
|
|
selected_chars_container = gr.Column(elem_id="selected_chars_container", visible=False)
|
|
|
|
with selected_chars_container:
|
|
gr.Markdown(("### 已选择的角色"))
|
|
selected_chars_rows = []
|
|
for i in range(5):
|
|
with gr.Row() as row:
|
|
name = gr.Textbox(label=("名称"), interactive=False, max_lines=1)
|
|
emotion = gr.Textbox(label=("情绪"), interactive=False, max_lines=1)
|
|
delete_btn = gr.Button(("删除"), scale=0)
|
|
selected_chars_rows.append((name, emotion, delete_btn, row))
|
|
|
|
|
|
|
|
|
|
lang.change(
|
|
fn=update_all_characters,
|
|
inputs=[lang, all_characters_state],
|
|
outputs=[all_characters_state, characters, gallery]
|
|
)
|
|
|
|
add_voice_button.click(
|
|
fn=add_new_voice,
|
|
inputs=[current_character, selected_characters, kind, lang, all_characters_state],
|
|
outputs=[current_character, character_name, info_type, selected_characters,
|
|
characters, gallery, selected_chars_container, all_characters_state]
|
|
).then(
|
|
fn=update_selected_chars_display,
|
|
inputs=[selected_characters],
|
|
outputs=[item for row in selected_chars_rows for item in row]
|
|
)
|
|
|
|
|
|
gallery.select(
|
|
fn=on_select,
|
|
inputs=[characters, selected_characters, all_characters_state],
|
|
outputs=[character_name, info_type, current_character, selected_characters]
|
|
)
|
|
|
|
info_type.change(
|
|
fn=update_character_info,
|
|
inputs=[character_name, info_type, current_character, all_characters_state],
|
|
outputs=[current_character, all_characters_state]
|
|
)
|
|
|
|
for i, (_, _, delete_btn, _) in enumerate(selected_chars_rows):
|
|
delete_btn.click(
|
|
fn=remove_character,
|
|
inputs=[gr.Number(value=i, visible=False), selected_characters],
|
|
outputs=[selected_characters, selected_chars_container]
|
|
).then(
|
|
fn=update_selected_chars_display,
|
|
inputs=[selected_characters],
|
|
outputs=[item for row in selected_chars_rows for item in row]
|
|
)
|
|
|
|
kind.change(
|
|
fn=update_gallery,
|
|
inputs=[kind, query, all_characters_state],
|
|
outputs=[characters, gallery, all_characters_state]
|
|
)
|
|
|
|
query.change(
|
|
fn=update_gallery,
|
|
inputs=[kind, query, all_characters_state],
|
|
outputs=[characters, gallery, all_characters_state]
|
|
)
|
|
|
|
with gr.Row():
|
|
with gr.Column():
|
|
text = gr.Textbox(label=("需要合成的文本"), value="", lines=10, max_lines=10)
|
|
inference_button = gr.Button(("🎉 合成语音 🎉"), variant="primary", size='lg')
|
|
with gr.Column():
|
|
output = gr.Audio(label=("输出的语音"), interactive=False, type="numpy")
|
|
cost_time = gr.Textbox(label=("合成时间"), interactive=False, show_label=False, max_lines=1)
|
|
try:
|
|
inference_button.click(
|
|
fn=generate,
|
|
inputs=[current_character, selected_characters, text, lang],
|
|
outputs=[output, cost_time],
|
|
)
|
|
except gr.Error as e:
|
|
gr.Error(e)
|
|
except Exception as e:
|
|
pass
|
|
gr.Markdown((terms))
|
|
|
|
if __name__ == '__main__':
|
|
demo.queue(default_concurrency_limit=8).launch(
|
|
server_name="0.0.0.0",
|
|
server_port=80,
|
|
show_api=False
|
|
)
|
|
|
