import discord import sys import soundfile as sf import numpy as np from tools.i18n.i18n import I18nAuto from GPT_SoVITS.test_webui import change_gpt_weights, change_sovits_weights, get_tts_wav import os from discord.ext import commands import asyncio # 創建 intents 實例 intents = discord.Intents.default() intents.messages = True # 初始化 Bot,並傳入 intents client = commands.Bot(command_prefix='!', intents=intents) i18n = I18nAuto() def remove_silence(audio_data, threshold=0.01): audio_data = np.array(audio_data) non_silent_indices = np.where(np.abs(audio_data) > threshold)[0] if non_silent_indices.size > 0: start_index = non_silent_indices[0] end_index = non_silent_indices[-1] + 1 return audio_data[start_index:end_index] return audio_data async def synthesize_audio(gpt_model_path, sovits_model_path, ref_audio_path, ref_text, target_text, output_path, language_combobox, language_combobox_02): await asyncio.to_thread(change_gpt_weights, gpt_path=gpt_model_path) await asyncio.to_thread(change_sovits_weights, sovits_path=sovits_model_path) synthesis_result = await asyncio.to_thread(get_tts_wav, ref_wav_path=ref_audio_path, prompt_text=ref_text, prompt_language=language_combobox, text=target_text, text_language=language_combobox_02) result_list = list(synthesis_result) if result_list: last_sampling_rate, last_audio_data = result_list[-1] last_audio_data = remove_silence(last_audio_data) if not os.path.exists(output_path): os.makedirs(output_path) output_wav_path = os.path.join(output_path, "output.wav") sf.write(output_wav_path, last_audio_data, last_sampling_rate) return output_wav_path return None @client.command(name="say") async def synthesize(ctx, *, target_text: str): ref_text = "幽魂盯上你了,進入射擊位置" output_path = 'GPT_SoVITS/output' gpt_model_path = 'GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt' sovits_model_path = 'GPT_SoVITS/pretrained_models/s2G488k.pth' ref_audio_path = 'GPT_SoVITS/pretrained_models/Ana.wav' language_combobox = i18n("中文") language_combobox_02 = i18n("多语种混合") output_wav_path = await synthesize_audio(gpt_model_path, sovits_model_path, ref_audio_path, ref_text, target_text, output_path, language_combobox, language_combobox_02) if output_wav_path: if ctx.author.voice: channel = ctx.author.voice.channel voice_client = await channel.connect() voice_client.play(discord.FFmpegPCMAudio(output_wav_path), after=lambda e: print('播放完成', e)) await ctx.send("正在播放生成的音頻...") # 等待音頻播放完成後斷開連接 while voice_client.is_playing(): await asyncio.sleep(1) await voice_client.disconnect() else: await ctx.send("請先進入語音頻道!") else: await ctx.send("合成失敗!")