Spaces:
Build error
Build error
Upload app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,615 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from openai import OpenAI
|
| 3 |
+
import requests
|
| 4 |
+
import json
|
| 5 |
+
# from volcenginesdkarkruntime import Ark
|
| 6 |
+
import torch
|
| 7 |
+
import torchaudio
|
| 8 |
+
from einops import rearrange
|
| 9 |
+
import argparse
|
| 10 |
+
import json
|
| 11 |
+
import os
|
| 12 |
+
import spaces
|
| 13 |
+
from tqdm import tqdm
|
| 14 |
+
import random
|
| 15 |
+
import numpy as np
|
| 16 |
+
import sys
|
| 17 |
+
import base64
|
| 18 |
+
|
| 19 |
+
from diffrhythm.infer.infer_utils import (
|
| 20 |
+
get_reference_latent,
|
| 21 |
+
get_lrc_token,
|
| 22 |
+
get_audio_style_prompt,
|
| 23 |
+
get_text_style_prompt,
|
| 24 |
+
prepare_model,
|
| 25 |
+
get_negative_style_prompt
|
| 26 |
+
)
|
| 27 |
+
from diffrhythm.infer.infer import inference
|
| 28 |
+
|
| 29 |
+
MAX_SEED = np.iinfo(np.int32).max
|
| 30 |
+
device=device = "cpu"
|
| 31 |
+
cfm, cfm_full, tokenizer, muq, vae = prepare_model(device)
|
| 32 |
+
cfm = torch.compile(cfm)
|
| 33 |
+
cfm_full = torch.compile(cfm_full)
|
| 34 |
+
|
| 35 |
+
@spaces.GPU(duration=40)
|
| 36 |
+
def infer_music(lrc, ref_audio_path, text_prompt, current_prompt_type, seed=42, randomize_seed=False, steps=32, cfg_strength=4.0, file_type='wav', odeint_method='euler', Music_Duration='95s', device="cpu" ):
|
| 37 |
+
# Ensure that inputs like ref_audio_path are processed as tensors
|
| 38 |
+
if isinstance(ref_audio_path, np.ndarray):
|
| 39 |
+
ref_audio_path = torch.tensor(ref_audio_path, dtype=torch.float32) # Ensure float32
|
| 40 |
+
|
| 41 |
+
# Ensure that any other tensors are consistent with the correct dtype
|
| 42 |
+
if isinstance(lrc, np.ndarray):
|
| 43 |
+
lrc = torch.tensor(lrc, dtype=torch.float32)
|
| 44 |
+
|
| 45 |
+
if Music_Duration == '95s':
|
| 46 |
+
max_frames = 2048
|
| 47 |
+
cfm_model = cfm
|
| 48 |
+
else:
|
| 49 |
+
max_frames = 6144
|
| 50 |
+
cfm_model = cfm_full
|
| 51 |
+
if randomize_seed:
|
| 52 |
+
seed = random.randint(0, MAX_SEED)
|
| 53 |
+
torch.manual_seed(seed)
|
| 54 |
+
sway_sampling_coef = -1 if steps < 32 else None
|
| 55 |
+
vocal_flag = False
|
| 56 |
+
try:
|
| 57 |
+
lrc_prompt, start_time = get_lrc_token(max_frames, lrc, tokenizer, device)
|
| 58 |
+
if current_prompt_type == 'audio':
|
| 59 |
+
style_prompt, vocal_flag = get_audio_style_prompt(muq, ref_audio_path)
|
| 60 |
+
else:
|
| 61 |
+
style_prompt = get_text_style_prompt(muq, text_prompt)
|
| 62 |
+
except Exception as e:
|
| 63 |
+
print(f"Error details: {e}")
|
| 64 |
+
raise gr.Error(f"Error: {str(e)}")
|
| 65 |
+
negative_style_prompt = get_negative_style_prompt(device)
|
| 66 |
+
latent_prompt = get_reference_latent(device, max_frames)
|
| 67 |
+
generated_song = inference(cfm_model=cfm_model,
|
| 68 |
+
vae_model=vae,
|
| 69 |
+
cond=latent_prompt,
|
| 70 |
+
text=lrc_prompt,
|
| 71 |
+
duration=max_frames,
|
| 72 |
+
style_prompt=style_prompt,
|
| 73 |
+
negative_style_prompt=negative_style_prompt,
|
| 74 |
+
steps=steps,
|
| 75 |
+
cfg_strength=cfg_strength,
|
| 76 |
+
sway_sampling_coef=sway_sampling_coef,
|
| 77 |
+
start_time=start_time,
|
| 78 |
+
file_type=file_type,
|
| 79 |
+
vocal_flag=vocal_flag,
|
| 80 |
+
odeint_method=odeint_method,
|
| 81 |
+
)
|
| 82 |
+
return generated_song
|
| 83 |
+
os.environ["PATH"] = "/opt/homebrew/bin:" + os.environ["PATH"]
|
| 84 |
+
def R1_infer1(theme, tags_gen, language):
|
| 85 |
+
try:
|
| 86 |
+
client = OpenAI(api_key=os.getenv('HS_DP_API'), base_url = "https://ark.cn-beijing.volces.com/api/v3")
|
| 87 |
+
|
| 88 |
+
llm_prompt = """
|
| 89 |
+
请围绕"{theme}"主题生成一首符合"{tags}"风格的语言为{language}的完整歌词。严格遵循以下要求:
|
| 90 |
+
|
| 91 |
+
### **强制格式规则**
|
| 92 |
+
1. **仅输出时间戳和歌词**,禁止任何括号、旁白、段落标记(如副歌、间奏、尾奏等注释)。
|
| 93 |
+
2. 每行格式必须为 `[mm:ss.xx]歌词内容`,时间戳与歌词间无空格,歌词内容需完整连贯。
|
| 94 |
+
3. 时间戳需自然分布,**第一句歌词起始时间不得为 [00:00.00]**,需考虑前奏空白。
|
| 95 |
+
|
| 96 |
+
### **内容与结构要求**
|
| 97 |
+
1. 歌词应富有变化,使情绪递进,整体连贯有层次感。**每行歌词长度应自然变化**,切勿长度一致,导致很格式化。
|
| 98 |
+
2. **时间戳分配应根据歌曲的标签、歌词的情感、节奏来合理推测**,而非机械地按照歌词长度分配。
|
| 99 |
+
3. 间奏/尾奏仅通过时间空白体现(如从 [02:30.00] 直接跳至 [02:50.00]),**无需文字描述**。
|
| 100 |
+
|
| 101 |
+
### **负面示例(禁止出现)**
|
| 102 |
+
- 错误:[01:30.00](钢琴间奏)
|
| 103 |
+
- 错误:[02:00.00][副歌]
|
| 104 |
+
- 错误:空行、换行符、注释
|
| 105 |
+
"""
|
| 106 |
+
|
| 107 |
+
response = client.chat.completions.create(
|
| 108 |
+
model="ep-20250304144033-nr9wl",
|
| 109 |
+
messages=[
|
| 110 |
+
{"role": "system", "content": "You are a professional musician who has been invited to make music-related comments."},
|
| 111 |
+
{"role": "user", "content": llm_prompt.format(theme=theme, tags=tags_gen, language=language)},
|
| 112 |
+
],
|
| 113 |
+
stream=False
|
| 114 |
+
)
|
| 115 |
+
|
| 116 |
+
info = response.choices[0].message.content
|
| 117 |
+
|
| 118 |
+
return info
|
| 119 |
+
|
| 120 |
+
except requests.exceptions.RequestException as e:
|
| 121 |
+
print(f'请求出错: {e}')
|
| 122 |
+
return {}
|
| 123 |
+
|
| 124 |
+
|
| 125 |
+
|
| 126 |
+
def R1_infer2(tags_lyrics, lyrics_input):
|
| 127 |
+
client = OpenAI(api_key=os.getenv('HS_DP_API'), base_url = "https://ark.cn-beijing.volces.com/api/v3")
|
| 128 |
+
|
| 129 |
+
llm_prompt = """
|
| 130 |
+
{lyrics_input}这是一首歌的歌词,每一行是一句歌词,{tags_lyrics}是我希望这首歌的风格,我现在想要给这首歌的每一句歌词打时间戳得到LRC,我希望时间戳分配应根据歌曲的标签、歌词的情感、节奏来合理推测,而非机械地按照歌词长度分配。第一句歌词的时间戳应考虑前奏长度,避免歌词从 `[00:00.00]` 直接开始。严格按照 LRC 格式输出歌词,每行格式为 `[mm:ss.xx]歌词内容`。最后的结果只输出LRC,不需要其他的解释。
|
| 131 |
+
"""
|
| 132 |
+
|
| 133 |
+
response = client.chat.completions.create(
|
| 134 |
+
model="ep-20250304144033-nr9wl",
|
| 135 |
+
messages=[
|
| 136 |
+
{"role": "system", "content": "You are a professional musician who has been invited to make music-related comments."},
|
| 137 |
+
{"role": "user", "content": llm_prompt.format(lyrics_input=lyrics_input, tags_lyrics=tags_lyrics)},
|
| 138 |
+
],
|
| 139 |
+
stream=False
|
| 140 |
+
)
|
| 141 |
+
|
| 142 |
+
info = response.choices[0].message.content
|
| 143 |
+
|
| 144 |
+
return info
|
| 145 |
+
|
| 146 |
+
css = """
|
| 147 |
+
/* Modern music-themed color palette */
|
| 148 |
+
:root {
|
| 149 |
+
--primary-pink: #ff69b4;
|
| 150 |
+
--secondary-pink: #ffb6c1;
|
| 151 |
+
--dark-purple: #4a154b;
|
| 152 |
+
--light-purple: #9c27b0;
|
| 153 |
+
--background: #1a1a1a;
|
| 154 |
+
--text-light: #ffffff;
|
| 155 |
+
--text-dark: #333333;
|
| 156 |
+
}
|
| 157 |
+
|
| 158 |
+
/* Global styles */
|
| 159 |
+
.gradio-container {
|
| 160 |
+
background: linear-gradient(135deg, #1a1a1a 0%, #2d1f3d 100%);
|
| 161 |
+
color: var(--text-light);
|
| 162 |
+
}
|
| 163 |
+
|
| 164 |
+
/* Enhanced text area styling */
|
| 165 |
+
.lyrics-scroll-box textarea {
|
| 166 |
+
height: 405px !important;
|
| 167 |
+
max-height: 500px !important;
|
| 168 |
+
overflow-y: auto !important;
|
| 169 |
+
white-space: pre-wrap;
|
| 170 |
+
line-height: 1.6;
|
| 171 |
+
padding: 15px !important;
|
| 172 |
+
border-radius: 12px !important;
|
| 173 |
+
border: 2px solid var(--primary-pink) !important;
|
| 174 |
+
background-color: rgba(26, 26, 26, 0.9) !important;
|
| 175 |
+
color: var(--text-light) !important;
|
| 176 |
+
font-family: 'Source Code Pro', monospace;
|
| 177 |
+
transition: all 0.3s ease;
|
| 178 |
+
box-shadow: 0 0 15px rgba(255, 105, 180, 0.2);
|
| 179 |
+
}
|
| 180 |
+
|
| 181 |
+
.lyrics-scroll-box textarea:focus {
|
| 182 |
+
border-color: var(--secondary-pink) !important;
|
| 183 |
+
box-shadow: 0 0 20px rgba(255, 105, 180, 0.4);
|
| 184 |
+
}
|
| 185 |
+
|
| 186 |
+
/* Music-themed button styling */
|
| 187 |
+
button.primary {
|
| 188 |
+
background: linear-gradient(45deg, var(--primary-pink), var(--light-purple)) !important;
|
| 189 |
+
border: none !important;
|
| 190 |
+
border-radius: 25px !important;
|
| 191 |
+
padding: 12px 24px !important;
|
| 192 |
+
color: white !important;
|
| 193 |
+
font-weight: 600 !important;
|
| 194 |
+
transition: all 0.3s ease;
|
| 195 |
+
text-shadow: 0 1px 2px rgba(0, 0, 0, 0.2);
|
| 196 |
+
box-shadow: 0 4px 15px rgba(255, 105, 180, 0.3);
|
| 197 |
+
}
|
| 198 |
+
|
| 199 |
+
button.primary:hover {
|
| 200 |
+
transform: translateY(-2px);
|
| 201 |
+
box-shadow: 0 6px 20px rgba(255, 105, 180, 0.5);
|
| 202 |
+
}
|
| 203 |
+
|
| 204 |
+
/* Tab styling */
|
| 205 |
+
.tabs {
|
| 206 |
+
border-radius: 12px;
|
| 207 |
+
overflow: hidden;
|
| 208 |
+
background: rgba(26, 26, 26, 0.8);
|
| 209 |
+
box-shadow: 0 4px 15px rgba(0, 0, 0, 0.2);
|
| 210 |
+
border: 1px solid rgba(255, 105, 180, 0.2);
|
| 211 |
+
}
|
| 212 |
+
|
| 213 |
+
.tab-selected {
|
| 214 |
+
background: linear-gradient(45deg, var(--primary-pink), var(--light-purple)) !important;
|
| 215 |
+
color: white !important;
|
| 216 |
+
border: none !important;
|
| 217 |
+
}
|
| 218 |
+
|
| 219 |
+
/* Examples section styling */
|
| 220 |
+
.gr-examples {
|
| 221 |
+
background: rgba(26, 26, 26, 0.8) !important;
|
| 222 |
+
border: 1px solid var(--primary-pink) !important;
|
| 223 |
+
border-radius: 12px;
|
| 224 |
+
margin: 1rem 0 !important;
|
| 225 |
+
padding: 1.5rem !important;
|
| 226 |
+
box-shadow: 0 4px 15px rgba(255, 105, 180, 0.2);
|
| 227 |
+
}
|
| 228 |
+
|
| 229 |
+
/* Slider improvements */
|
| 230 |
+
.gr-slider {
|
| 231 |
+
padding: 12px 0;
|
| 232 |
+
}
|
| 233 |
+
|
| 234 |
+
.gr-slider .handle {
|
| 235 |
+
background: var(--primary-pink) !important;
|
| 236 |
+
box-shadow: 0 0 10px var(--primary-pink);
|
| 237 |
+
}
|
| 238 |
+
|
| 239 |
+
.gr-slider .track {
|
| 240 |
+
background: rgba(255, 105, 180, 0.3) !important;
|
| 241 |
+
}
|
| 242 |
+
|
| 243 |
+
/* Accordion styling */
|
| 244 |
+
.gr-accordion {
|
| 245 |
+
border: 1px solid var(--primary-pink);
|
| 246 |
+
border-radius: 12px;
|
| 247 |
+
margin: 1rem 0;
|
| 248 |
+
background: rgba(26, 26, 26, 0.8);
|
| 249 |
+
}
|
| 250 |
+
|
| 251 |
+
.gr-accordion summary {
|
| 252 |
+
padding: 15px;
|
| 253 |
+
background: rgba(255, 105, 180, 0.1);
|
| 254 |
+
cursor: pointer;
|
| 255 |
+
color: var(--text-light);
|
| 256 |
+
}
|
| 257 |
+
|
| 258 |
+
/* Audio player styling */
|
| 259 |
+
.gr-audio {
|
| 260 |
+
border-radius: 12px;
|
| 261 |
+
overflow: hidden;
|
| 262 |
+
margin: 1rem 0;
|
| 263 |
+
background: rgba(26, 26, 26, 0.8);
|
| 264 |
+
border: 1px solid var(--primary-pink);
|
| 265 |
+
}
|
| 266 |
+
|
| 267 |
+
/* Radio button styling */
|
| 268 |
+
.gr-radio {
|
| 269 |
+
gap: 12px;
|
| 270 |
+
}
|
| 271 |
+
|
| 272 |
+
.gr-radio label {
|
| 273 |
+
padding: 10px 20px;
|
| 274 |
+
border-radius: 25px;
|
| 275 |
+
transition: all 0.3s ease;
|
| 276 |
+
background: rgba(255, 105, 180, 0.1);
|
| 277 |
+
color: var(--text-light);
|
| 278 |
+
}
|
| 279 |
+
|
| 280 |
+
.gr-radio label:hover {
|
| 281 |
+
background: rgba(255, 105, 180, 0.2);
|
| 282 |
+
}
|
| 283 |
+
|
| 284 |
+
.gr-radio input:checked + label {
|
| 285 |
+
background: linear-gradient(45deg, var(--primary-pink), var(--light-purple));
|
| 286 |
+
color: white;
|
| 287 |
+
}
|
| 288 |
+
|
| 289 |
+
/* Input field styling */
|
| 290 |
+
.gr-input {
|
| 291 |
+
background: rgba(26, 26, 26, 0.8) !important;
|
| 292 |
+
border: 2px solid var(--primary-pink) !important;
|
| 293 |
+
border-radius: 12px !important;
|
| 294 |
+
color: var(--text-light) !important;
|
| 295 |
+
padding: 12px !important;
|
| 296 |
+
}
|
| 297 |
+
|
| 298 |
+
.gr-input:focus {
|
| 299 |
+
box-shadow: 0 0 15px rgba(255, 105, 180, 0.3);
|
| 300 |
+
}
|
| 301 |
+
|
| 302 |
+
/* Dropdown styling */
|
| 303 |
+
.gr-dropdown {
|
| 304 |
+
background: rgba(26, 26, 26, 0.8) !important;
|
| 305 |
+
border: 2px solid var(--primary-pink) !important;
|
| 306 |
+
border-radius: 12px !important;
|
| 307 |
+
color: var(--text-light) !important;
|
| 308 |
+
}
|
| 309 |
+
|
| 310 |
+
/* Checkbox styling */
|
| 311 |
+
.gr-checkbox {
|
| 312 |
+
accent-color: var(--primary-pink);
|
| 313 |
+
}
|
| 314 |
+
|
| 315 |
+
/* Loading animation */
|
| 316 |
+
.progress-bar {
|
| 317 |
+
background: linear-gradient(45deg, var(--primary-pink), var(--light-purple));
|
| 318 |
+
}
|
| 319 |
+
|
| 320 |
+
/* Scrollbar styling */
|
| 321 |
+
::-webkit-scrollbar {
|
| 322 |
+
width: 10px;
|
| 323 |
+
}
|
| 324 |
+
|
| 325 |
+
::-webkit-scrollbar-track {
|
| 326 |
+
background: rgba(26, 26, 26, 0.8);
|
| 327 |
+
}
|
| 328 |
+
|
| 329 |
+
::-webkit-scrollbar-thumb {
|
| 330 |
+
background: var(--primary-pink);
|
| 331 |
+
border-radius: 5px;
|
| 332 |
+
}
|
| 333 |
+
|
| 334 |
+
/* Responsive improvements */
|
| 335 |
+
@media (max-width: 768px) {
|
| 336 |
+
.gr-examples {
|
| 337 |
+
padding: 1rem !important;
|
| 338 |
+
}
|
| 339 |
+
|
| 340 |
+
button.primary {
|
| 341 |
+
width: 100%;
|
| 342 |
+
}
|
| 343 |
+
}
|
| 344 |
+
"""
|
| 345 |
+
|
| 346 |
+
|
| 347 |
+
with gr.Blocks(css=css) as demo:
|
| 348 |
+
gr.HTML(f"""
|
| 349 |
+
<div style="text-align: center; margin-bottom: 2rem;">
|
| 350 |
+
# <div style="display: inline-block; background: rgba(26, 26, 26, 0.9); padding: 25px; border-radius: 20px; box-shadow: 0 0 30px rgba(255, 105, 180, 0.3);">
|
| 351 |
+
# # <img src=''
|
| 352 |
+
# # style='width: 200px; height: auto; margin-bottom: 1rem; border-radius: 15px; box-shadow: 0 4px 15px rgba(255, 105, 180, 0.4);'>
|
| 353 |
+
|
| 354 |
+
<h1 style="font-size: 2.8em; font-weight: bold; margin: 0.5rem 0; background: linear-gradient(45deg, #ff69b4, #9c27b0); -webkit-background-clip: text; -webkit-text-fill-color: transparent;">
|
| 355 |
+
TuneWeave (Generate Your Music)
|
| 356 |
+
</h1>
|
| 357 |
+
|
| 358 |
+
<div style="display: flex; gap: 15px; justify-content: center; margin-top: 1.5rem;">
|
| 359 |
+
<a href="https://arxiv.org/abs/2503.01183" style="text-decoration: none;">
|
| 360 |
+
<button style="background: linear-gradient(45deg, #ff69b4, #9c27b0); color: white; border: none; padding: 10px 20px; border-radius: 25px; cursor: pointer; font-weight: 600; transition: all 0.3s ease; box-shadow: 0 4px 15px rgba(255, 105, 180, 0.3);">
|
| 361 |
+
📄 Paper
|
| 362 |
+
</button>
|
| 363 |
+
</a>
|
| 364 |
+
<a href="https://github.com/ASLP-lab/DiffRhythm" style="text-decoration: none;">
|
| 365 |
+
<button style="background: linear-gradient(45deg, #ff69b4, #9c27b0); color: white; border: none; padding: 10px 20px; border-radius: 25px; cursor: pointer; font-weight: 600; transition: all 0.3s ease; box-shadow: 0 4px 15px rgba(255, 105, 180, 0.3);">
|
| 366 |
+
💻 GitHub
|
| 367 |
+
</button>
|
| 368 |
+
</a>
|
| 369 |
+
<a href="https://aslp-lab.github.io/DiffRhythm.github.io/" style="text-decoration: none;">
|
| 370 |
+
<button style="background: linear-gradient(45deg, #ff69b4, #9c27b0); color: white; border: none; padding: 10px 20px; border-radius: 25px; cursor: pointer; font-weight: 600; transition: all 0.3s ease; box-shadow: 0 4px 15px rgba(255, 105, 180, 0.3);">
|
| 371 |
+
🌐 Project
|
| 372 |
+
</button>
|
| 373 |
+
</a>
|
| 374 |
+
</div>
|
| 375 |
+
</div>
|
| 376 |
+
</div>
|
| 377 |
+
""")
|
| 378 |
+
|
| 379 |
+
with gr.Tabs() as tabs:
|
| 380 |
+
|
| 381 |
+
# page 1
|
| 382 |
+
with gr.Tab("Music Generate", id=0):
|
| 383 |
+
with gr.Row():
|
| 384 |
+
with gr.Column():
|
| 385 |
+
lrc = gr.Textbox(
|
| 386 |
+
label="Lyrics",
|
| 387 |
+
placeholder="Input the full lyrics",
|
| 388 |
+
lines=12,
|
| 389 |
+
max_lines=50,
|
| 390 |
+
elem_classes="lyrics-scroll-box",
|
| 391 |
+
value="""[00:04.34]Tell me that I'm special\n[00:06.57]Tell me I look pretty\n[00:08.46]Tell me I'm a little angel\n[00:10.58]Sweetheart of your city\n[00:13.64]Say what I'm dying to hear\n[00:17.35]Cause I'm dying to hear you\n[00:20.86]Tell me I'm that new thing\n[00:22.93]Tell me that I'm relevant\n[00:24.96]Tell me that I got a big heart\n[00:27.04]Then back it up with evidence\n[00:29.94]I need it and I don't know why\n[00:34.28]This late at night\n[00:36.32]Isn't it lonely\n[00:39.24]I'd do anything to make you want me\n[00:43.40]I'd give it all up if you told me\n[00:47.42]That I'd be\n[00:49.43]The number one girl in your eyes\n[00:52.85]Your one and only\n[00:55.74]So what's it gon' take for you to want me\n[00:59.78]I'd give it all up if you told me\n[01:03.89]That I'd be\n[01:05.94]The number one girl in your eyes\n[01:11.34]Tell me I'm going real big places\n[01:14.32]Down to earth so friendly\n[01:16.30]And even through all the phases\n[01:18.46]Tell me you accept me\n[01:21.56]Well that's all I'm dying to hear\n[01:25.30]Yeah I'm dying to hear you\n[01:28.91]Tell me that you need me\n[01:30.85]Tell me that I'm loved\n[01:32.90]Tell me that I'm worth it\n[01:34.95]And that I'm enough\n[01:37.91]I need it and I don't know why\n[01:42.08]This late at night\n[01:44.24]Isn't it lonely\n[01:47.18]I'd do anything to make you want me\n[01:51.30]I'd give it all up if you told me\n[01:55.32]That I'd be\n[01:57.35]The number one girl in your eyes\n[02:00.72]Your one and only\n[02:03.57]So what's it gon' take for you to want me\n[02:07.78]I'd give it all up if you told me\n[02:11.74]That I'd be\n[02:13.86]The number one girl in your eyes\n[02:17.03]The girl in your eyes\n[02:21.05]The girl in your eyes\n[02:26.30]Tell me I'm the number one girl\n[02:28.44]I'm the number one girl in your eyes\n[02:33.49]The girl in your eyes\n[02:37.58]The girl in your eyes\n[02:42.74]Tell me I'm the number one girl\n[02:44.88]I'm the number one girl in your eyes\n[02:49.91]Well isn't it lonely\n[02:53.19]I'd do anything to make you want me\n[02:57.10]I'd give it all up if you told me\n[03:01.15]That I'd be\n[03:03.31]The number one girl in your eyes\n[03:06.57]Your one and only\n[03:09.42]So what's it gon' take for you to want me\n[03:13.50]I'd give it all up if you told me\n[03:17.56]That I'd be\n[03:19.66]The number one girl in your eyes\n[03:25.74]The number one girl in your eyes"""
|
| 392 |
+
)
|
| 393 |
+
|
| 394 |
+
current_prompt_type = gr.State(value="audio")
|
| 395 |
+
with gr.Tabs() as inside_tabs:
|
| 396 |
+
with gr.Tab("Audio Prompt"):
|
| 397 |
+
audio_prompt = gr.Audio(label="Audio Prompt", type="filepath", value="./src/prompt/default.wav")
|
| 398 |
+
with gr.Tab("Text Prompt"):
|
| 399 |
+
text_prompt = gr.Textbox(
|
| 400 |
+
label="Text Prompt",
|
| 401 |
+
placeholder="Enter the Text Prompt, eg: emotional piano pop",
|
| 402 |
+
)
|
| 403 |
+
def update_prompt_type(evt: gr.SelectData):
|
| 404 |
+
return "audio" if evt.index == 0 else "text"
|
| 405 |
+
|
| 406 |
+
inside_tabs.select(
|
| 407 |
+
fn=update_prompt_type,
|
| 408 |
+
outputs=current_prompt_type
|
| 409 |
+
)
|
| 410 |
+
|
| 411 |
+
with gr.Column():
|
| 412 |
+
with gr.Accordion("Best Practices Guide", open=True):
|
| 413 |
+
gr.Markdown("""
|
| 414 |
+
1. **Lyrics Format Requirements**
|
| 415 |
+
- Each line must follow: `[mm:ss.xx]Lyric content`
|
| 416 |
+
- Example of valid format:
|
| 417 |
+
```
|
| 418 |
+
[00:10.00]Moonlight spills through broken blinds
|
| 419 |
+
[00:13.20]Your shadow dances on the dashboard shrine
|
| 420 |
+
```
|
| 421 |
+
|
| 422 |
+
2. **Audio Prompt Requirements**
|
| 423 |
+
- Reference audio should be ≥ 1 second, audio >10 seconds will be randomly clipped into 10 seconds
|
| 424 |
+
- For optimal results, the 10-second clips should be carefully selected
|
| 425 |
+
- Shorter clips may lead to incoherent generation
|
| 426 |
+
3. **Supported Languages**
|
| 427 |
+
- **Chinese and English**
|
| 428 |
+
- More languages comming soon
|
| 429 |
+
|
| 430 |
+
4. **Others**
|
| 431 |
+
- If loading audio result is slow, you can select Output Format as mp3 in Advanced Settings.
|
| 432 |
+
|
| 433 |
+
""")
|
| 434 |
+
Music_Duration = gr.Radio(["95s", "285s"], label="Music Duration", value="95s")
|
| 435 |
+
|
| 436 |
+
lyrics_btn = gr.Button("Generate", variant="primary")
|
| 437 |
+
audio_output = gr.Audio(label="Audio Result", type="filepath", elem_id="audio_output")
|
| 438 |
+
with gr.Accordion("Advanced Settings", open=False):
|
| 439 |
+
seed = gr.Slider(
|
| 440 |
+
label="Seed",
|
| 441 |
+
minimum=0,
|
| 442 |
+
maximum=MAX_SEED,
|
| 443 |
+
step=1,
|
| 444 |
+
value=0,
|
| 445 |
+
)
|
| 446 |
+
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
|
| 447 |
+
|
| 448 |
+
steps = gr.Slider(
|
| 449 |
+
minimum=10,
|
| 450 |
+
maximum=100,
|
| 451 |
+
value=32,
|
| 452 |
+
step=1,
|
| 453 |
+
label="Diffusion Steps",
|
| 454 |
+
interactive=True,
|
| 455 |
+
elem_id="step_slider"
|
| 456 |
+
)
|
| 457 |
+
cfg_strength = gr.Slider(
|
| 458 |
+
minimum=1,
|
| 459 |
+
maximum=10,
|
| 460 |
+
value=4.0,
|
| 461 |
+
step=0.5,
|
| 462 |
+
label="CFG Strength",
|
| 463 |
+
interactive=True,
|
| 464 |
+
elem_id="step_slider"
|
| 465 |
+
)
|
| 466 |
+
odeint_method = gr.Radio(["euler", "midpoint", "rk4","implicit_adams"], label="ODE Solver", value="euler")
|
| 467 |
+
file_type = gr.Dropdown(["wav", "mp3", "ogg"], label="Output Format", value="wav")
|
| 468 |
+
|
| 469 |
+
|
| 470 |
+
gr.Examples(
|
| 471 |
+
examples=[
|
| 472 |
+
["./src/prompt/pop_cn.wav"],
|
| 473 |
+
["./src/prompt/pop_en.wav"],
|
| 474 |
+
["./src/prompt/rock_cn.wav"],
|
| 475 |
+
["./src/prompt/rock_en.wav"],
|
| 476 |
+
["./src/prompt/country_cn.wav"],
|
| 477 |
+
["./src/prompt/country_en.wav"],
|
| 478 |
+
["./src/prompt/classic_cn.wav"],
|
| 479 |
+
["./src/prompt/classic_en.wav"],
|
| 480 |
+
["./src/prompt/jazz_cn.wav"],
|
| 481 |
+
["./src/prompt/jazz_en.wav"],
|
| 482 |
+
["./src/prompt/rap_cn.wav"],
|
| 483 |
+
["./src/prompt/rap_en.wav"],
|
| 484 |
+
["./src/prompt/default.wav"]
|
| 485 |
+
],
|
| 486 |
+
inputs=[audio_prompt],
|
| 487 |
+
label="Audio Examples",
|
| 488 |
+
examples_per_page=13,
|
| 489 |
+
elem_id="audio-examples-container"
|
| 490 |
+
)
|
| 491 |
+
|
| 492 |
+
gr.Examples(
|
| 493 |
+
examples=[
|
| 494 |
+
["Pop Emotional Piano"],
|
| 495 |
+
["流行 情感 钢琴"],
|
| 496 |
+
["Indie folk ballad, coming-of-age themes, acoustic guitar picking with harmonica interludes"],
|
| 497 |
+
["独立民谣, 成长主题, 原声吉他弹奏与口琴间奏"]
|
| 498 |
+
],
|
| 499 |
+
inputs=[text_prompt],
|
| 500 |
+
label="Text Examples",
|
| 501 |
+
examples_per_page=4,
|
| 502 |
+
elem_id="text-examples-container"
|
| 503 |
+
)
|
| 504 |
+
|
| 505 |
+
gr.Examples(
|
| 506 |
+
examples=[
|
| 507 |
+
["""[00:04.34]Tell me that I'm special\n[00:06.57]Tell me I look pretty\n[00:08.46]Tell me I'm a little angel\n[00:10.58]Sweetheart of your city\n[00:13.64]Say what I'm dying to hear\n[00:17.35]Cause I'm dying to hear you\n[00:20.86]Tell me I'm that new thing\n[00:22.93]Tell me that I'm relevant\n[00:24.96]Tell me that I got a big heart\n[00:27.04]Then back it up with evidence\n[00:29.94]I need it and I don't know why\n[00:34.28]This late at night\n[00:36.32]Isn't it lonely\n[00:39.24]I'd do anything to make you want me\n[00:43.40]I'd give it all up if you told me\n[00:47.42]That I'd be\n[00:49.43]The number one girl in your eyes\n[00:52.85]Your one and only\n[00:55.74]So what's it gon' take for you to want me\n[00:59.78]I'd give it all up if you told me\n[01:03.89]That I'd be\n[01:05.94]The number one girl in your eyes\n[01:11.34]Tell me I'm going real big places\n[01:14.32]Down to earth so friendly\n[01:16.30]And even through all the phases\n[01:18.46]Tell me you accept me\n[01:21.56]Well that's all I'm dying to hear\n[01:25.30]Yeah I'm dying to hear you\n[01:28.91]Tell me that you need me\n[01:30.85]Tell me that I'm loved\n[01:32.90]Tell me that I'm worth it\n[01:34.95]And that I'm enough\n[01:37.91]I need it and I don't know why\n[01:42.08]This late at night\n[01:44.24]Isn't it lonely\n[01:47.18]I'd do anything to make you want me\n[01:51.30]I'd give it all up if you told me\n[01:55.32]That I'd be\n[01:57.35]The number one girl in your eyes\n[02:00.72]Your one and only\n[02:03.57]So what's it gon' take for you to want me\n[02:07.78]I'd give it all up if you told me\n[02:11.74]That I'd be\n[02:13.86]The number one girl in your eyes\n[02:17.03]The girl in your eyes\n[02:21.05]The girl in your eyes\n[02:26.30]Tell me I'm the number one girl\n[02:28.44]I'm the number one girl in your eyes\n[02:33.49]The girl in your eyes\n[02:37.58]The girl in your eyes\n[02:42.74]Tell me I'm the number one girl\n[02:44.88]I'm the number one girl in your eyes\n[02:49.91]Well isn't it lonely\n[02:53.19]I'd do anything to make you want me\n[02:57.10]I'd give it all up if you told me\n[03:01.15]That I'd be\n[03:03.31]The number one girl in your eyes\n[03:06.57]Your one and only\n[03:09.42]So what's it gon' take for you to want me\n[03:13.50]I'd give it all up if you told me\n[03:17.56]That I'd be\n[03:19.66]The number one girl in your eyes\n[03:25.74]The number one girl in your eyes"""],
|
| 508 |
+
["""[00:00.52]Abracadabra abracadabra\n[00:03.97]Ha\n[00:04.66]Abracadabra abracadabra\n[00:12.02]Yeah\n[00:15.80]Pay the toll to the angels\n[00:19.08]Drawin' circles in the clouds\n[00:23.31]Keep your mind on the distance\n[00:26.67]When the devil turns around\n[00:30.95]Hold me in your heart tonight\n[00:34.11]In the magic of the dark moonlight\n[00:38.44]Save me from this empty fight\n[00:43.83]In the game of life\n[00:45.84]Like a poem said by a lady in red\n[00:49.45]You hear the last few words of your life\n[00:53.15]With a haunting dance now you're both in a trance\n[00:56.90]It's time to cast your spell on the night\n[01:01.40]Abracadabra ama-ooh-na-na\n[01:04.88]Abracadabra porta-ooh-ga-ga\n[01:08.92]Abracadabra abra-ooh-na-na\n[01:12.30]In her tongue she's sayin'\n[01:14.76]Death or love tonight\n[01:18.61]Abracadabra abracadabra\n[01:22.18]Abracadabra abracadabra\n[01:26.08]Feel the beat under your feet\n[01:27.82]The floor's on fire\n[01:29.90]Abracadabra abracadabra\n[01:33.78]Choose the road on the west side\n[01:37.09]As the dust flies watch it burn\n[01:41.45]Don't waste time on feeling\n[01:44.64]Your depression won't return\n[01:49.15]Hold me in your heart tonight\n[01:52.21]In the magic of the dark moonlight\n[01:56.54]Save me from this empty fight\n[02:01.77]In the game of life\n[02:03.94]Like a poem said by a lady in red\n[02:07.52]You hear the last few words of your life\n[02:11.19]With a haunting dance now you're both in a trance\n[02:14.95]It's time to cast your spell on the night\n[02:19.53]Abracadabra ama-ooh-na-na\n[02:22.71]Abracadabra porta-ooh-ga-ga\n[02:26.94]Abracadabra abra-ooh-na-na\n[02:30.42]In her tongue she's sayin'\n[02:32.83]Death or love tonight\n[02:36.55]Abracadabra abracadabra\n[02:40.27]Abracadabra abracadabra\n[02:44.19]Feel the beat under your feet\n[02:46.14]The floor's on fire\n[02:47.95]Abracadabra abracadabra\n[02:51.17]Phantom of the dance floor come to me\n[02:58.46]Sing for me a sinful melody\n[03:06.51]Ah-ah-ah-ah-ah ah-ah ah-ah\n[03:13.76]Ah-ah-ah-ah-ah ah-ah ah-ah\n[03:22.39]Abracadabra ama-ooh-na-na\n[03:25.66]Abracadabra porta-ooh-ga-ga\n[03:29.87]Abracadabra abra-ooh-na-na\n[03:33.16]In her tongue she's sayin'\n[03:35.55]Death or love tonight"""],
|
| 509 |
+
# ["""[00:00.27]只因你太美 baby 只因你太美 baby\n[00:08.95]只因你实在是太美 baby\n[00:13.99]只因你太美 baby\n[00:18.89]迎面走来的你让我如此蠢蠢欲动\n[00:20.88]这种感觉我从未有\n[00:21.79]Cause I got a crush on you who you\n[00:25.74]你是我的我是你的谁\n[00:28.09]再多一眼看一眼就会爆炸\n[00:30.31]再近一点靠近点快被融化\n[00:32.49]想要把你占为己有 baby bae\n[00:34.60]不管走到哪里\n[00:35.44]都会想起的人是你 you you\n[00:38.12]我应该拿你怎样\n[00:39.61]Uh 所有人都在看着你\n[00:42.36]我的心总是不安\n[00:44.18]Oh 我现在已病入膏肓\n[00:46.63]Eh oh\n[00:47.84]难道真的因你而疯狂吗\n[00:51.57]我本来不是这种人\n[00:53.59]因你变成奇怪的人\n[00:55.77]第一次呀变成这样的我\n[01:01.23]不管我怎么去否认\n[01:03.21]只因你太美 baby 只因你太美 baby\n[01:11.46]只因你实在是太美 baby\n[01:16.75]只因你太美 baby\n[01:21.09]Oh eh oh\n[01:22.82]现在确认地告诉我\n[01:25.26]Oh eh oh\n[01:27.31]你到底属于谁\n[01:29.98]Oh eh oh\n[01:31.70]现在确认地告诉我\n[01:34.45]Oh eh oh\n[01:36.35]你到底属于谁\n[01:37.65]就是现在告诉我\n[01:40.00]跟着那节奏 缓缓 make wave\n[01:42.42]甜蜜的奶油 it's your birthday cake\n[01:44.66]男人们的 game call me 你恋人\n[01:46.83]别被欺骗愉快的 I wanna play\n[01:48.83]我的脑海每分每秒为你一人沉醉\n[01:50.90]最迷人让我神魂颠倒是你身上香水\n[01:53.30]Oh right baby I'm fall in love with you\n[01:55.20]我的一切你都拿走\n[01:56.40]只要有你就已足够\n[01:58.56]我到底应该怎样\n[02:00.37]Uh 我心里一直很不安\n[02:03.12]其他男人们的视线\n[02:04.84]Oh 全都只看着你的脸\n[02:07.33]Eh oh\n[02:08.39]难道真的因你而疯狂吗\n[02:12.43]我本来不是这种人\n[02:14.35]因你变成奇怪的人\n[02:16.59]第一次呀变成这样的我\n[02:21.76]不管我怎么去否认\n[02:24.03]只因你太美 baby 只因你太美 baby\n[02:32.37]只因你实在是太美 baby\n[02:37.49]只因你太美 baby\n[02:43.66]我愿意把我的全部都给你\n[02:47.19]我每天在梦里都梦见你\n[02:49.13]还有我闭着眼睛也能看到你\n[02:52.58]现在开始我只准你看我\n[02:56.28]I don't wanna wake up in dream\n[02:57.92]我只想看你这是真心话\n[02:59.86]只因你太美 baby 只因你太美 baby\n[03:08.20]只因你实在是太美 baby\n[03:13.22]只因你太美 baby\n[03:17.69]Oh eh oh\n[03:19.36]现在确认的告诉我\n[03:21.91]Oh eh oh\n[03:23.85]你到底属于谁\n[03:26.58]Oh eh oh\n[03:28.32]现在确认的告诉我\n[03:30.95]Oh eh oh\n[03:32.82]你到底属于谁就是现在告诉我"""]
|
| 510 |
+
],
|
| 511 |
+
|
| 512 |
+
inputs=[lrc],
|
| 513 |
+
label="Lrc Examples",
|
| 514 |
+
examples_per_page=3,
|
| 515 |
+
elem_id="lrc-examples-container",
|
| 516 |
+
)
|
| 517 |
+
|
| 518 |
+
|
| 519 |
+
# page 2
|
| 520 |
+
with gr.Tab("Lyrics Generate", id=1):
|
| 521 |
+
with gr.Row():
|
| 522 |
+
with gr.Column():
|
| 523 |
+
with gr.Accordion("Notice", open=False):
|
| 524 |
+
gr.Markdown("**Two Generation Modes:**\n1. Generate from theme & tags\n2. Add timestamps to existing lyrics")
|
| 525 |
+
|
| 526 |
+
with gr.Group():
|
| 527 |
+
gr.Markdown("### Method 1: Generate from Theme")
|
| 528 |
+
theme = gr.Textbox(label="theme", placeholder="Enter song theme, e.g: Love and Heartbreak")
|
| 529 |
+
tags_gen = gr.Textbox(label="tags", placeholder="Enter song tags, e.g: pop confidence healing")
|
| 530 |
+
language = gr.Radio(["cn", "en"], label="Language", value="en")
|
| 531 |
+
gen_from_theme_btn = gr.Button("Generate LRC (From Theme)", variant="primary")
|
| 532 |
+
|
| 533 |
+
gr.Examples(
|
| 534 |
+
examples=[
|
| 535 |
+
[
|
| 536 |
+
"Love and Heartbreak",
|
| 537 |
+
"vocal emotional piano pop",
|
| 538 |
+
"en"
|
| 539 |
+
],
|
| 540 |
+
[
|
| 541 |
+
"Heroic Epic",
|
| 542 |
+
"choir orchestral powerful",
|
| 543 |
+
"cn"
|
| 544 |
+
]
|
| 545 |
+
],
|
| 546 |
+
inputs=[theme, tags_gen, language],
|
| 547 |
+
label="Examples: Generate from Theme"
|
| 548 |
+
)
|
| 549 |
+
|
| 550 |
+
with gr.Group(visible=True):
|
| 551 |
+
gr.Markdown("### Method 2: Add Timestamps to Lyrics")
|
| 552 |
+
tags_lyrics = gr.Textbox(label="tags", placeholder="Enter song tags, e.g: ballad piano slow")
|
| 553 |
+
lyrics_input = gr.Textbox(
|
| 554 |
+
label="Raw Lyrics (without timestamps)",
|
| 555 |
+
placeholder="Enter plain lyrics (without timestamps), e.g:\nYesterday\nAll my troubles...",
|
| 556 |
+
lines=10,
|
| 557 |
+
max_lines=50,
|
| 558 |
+
elem_classes="lyrics-scroll-box"
|
| 559 |
+
)
|
| 560 |
+
|
| 561 |
+
gen_from_lyrics_btn = gr.Button("Generate LRC (From Lyrics)", variant="primary")
|
| 562 |
+
|
| 563 |
+
gr.Examples(
|
| 564 |
+
examples=[
|
| 565 |
+
[
|
| 566 |
+
"acoustic folk happy",
|
| 567 |
+
"""I'm sitting here in the boring room\nIt's just another rainy Sunday afternoon"""
|
| 568 |
+
],
|
| 569 |
+
[
|
| 570 |
+
"electronic dance energetic",
|
| 571 |
+
"""We're living in a material world\nAnd I am a material girl"""
|
| 572 |
+
]
|
| 573 |
+
],
|
| 574 |
+
inputs=[tags_lyrics, lyrics_input],
|
| 575 |
+
label="Examples: Generate from Lyrics"
|
| 576 |
+
)
|
| 577 |
+
|
| 578 |
+
|
| 579 |
+
with gr.Column():
|
| 580 |
+
lrc_output = gr.Textbox(
|
| 581 |
+
label="Generated LRC",
|
| 582 |
+
placeholder="Timed lyrics will appear here",
|
| 583 |
+
lines=57,
|
| 584 |
+
elem_classes="lrc-output",
|
| 585 |
+
show_copy_button=True
|
| 586 |
+
)
|
| 587 |
+
|
| 588 |
+
# Bind functions
|
| 589 |
+
gen_from_theme_btn.click(
|
| 590 |
+
fn=R1_infer1,
|
| 591 |
+
inputs=[theme, tags_gen, language],
|
| 592 |
+
outputs=lrc_output
|
| 593 |
+
)
|
| 594 |
+
|
| 595 |
+
gen_from_lyrics_btn.click(
|
| 596 |
+
fn=R1_infer2,
|
| 597 |
+
inputs=[tags_lyrics, lyrics_input],
|
| 598 |
+
outputs=lrc_output
|
| 599 |
+
)
|
| 600 |
+
|
| 601 |
+
tabs.select(
|
| 602 |
+
lambda s: None,
|
| 603 |
+
None,
|
| 604 |
+
None
|
| 605 |
+
)
|
| 606 |
+
|
| 607 |
+
lyrics_btn.click(
|
| 608 |
+
fn=infer_music,
|
| 609 |
+
inputs=[lrc, audio_prompt, text_prompt, current_prompt_type, seed, randomize_seed, steps, cfg_strength, file_type, odeint_method, Music_Duration],
|
| 610 |
+
outputs=audio_output
|
| 611 |
+
)
|
| 612 |
+
|
| 613 |
+
|
| 614 |
+
if __name__ == "__main__":
|
| 615 |
+
demo.launch(share=True)
|