speak.py ダウンロード/コピー
speak.py
speak.py
1# このプログラムの実行には以下のライブラリと外部依存が必要です。
2# pip install chardet pyttsx3 openai google-genai edge-tts pydub pyperclip
3#
4# 外部依存:
5# - ffmpeg.exe: 環境PATHに設定するか、pydubに検出させる必要があります。
6# - AquesTalkPlayer.exe: Windowsでのみ使用可能。実行パスを指定する必要があります。
7# - OpenAI API Key: 環境変数 OPENAI_API_KEY に設定する必要があります。
8# - Gemini API Key: 環境変数 GEMINI_API_KEY に設定する必要があります。
9
10import os
11import sys
12import argparse
13import traceback
14
15missing = []
16for lib in ["chardet", "pyperclip", "tktts"]:
17 try:
18 __import__(lib)
19 except ImportError:
20# missing.append(lib)
21 traceback.print_exc()
22 pass
23
24if missing:
25 print(f"Error: Missing libraries:\n{', '.join(missing)}")
26 print(" install: pip chardet")
27 input("\nPress ENTER to terminate>>\n")
28 sys.exit(1)
29
30import chardet
31import pyperclip
32
33try:
34 import tktts
35 from tktts import tkTTS
36except Exception as e:
37 print(f"\nWarning in tktts.py: Import error for tktts_pyttsx3")
38 print("------------------------------------------------------------------")
39 print(f"Error message: {e}")
40 print("Traceback:")
41 traceback.print_exc()
42 print("------------------------------------------------------------------")
43
44
45DEFAULT_INPUT = "clip"
46DEFAULT_ENGINE = "pyttsx3"
47DEFAULT_VOICEVOX_ENDPOINT = "http://127.0.0.1:50021"
48DEFAULT_AQUESTALK_PATH = "AquesTalkPlayer.exe"
49DEFAULT_TEMP_DIR = "tts_temp_wavs"
50
51VOICE_MAPS = {
52 "pyttsx3": {"四国めたん": "Zira", "ずんだもん": "David", "れいむ": "Zira", "まりさ": "David"},
53 "qwen3": {"四国めたん": "Ono_Anna", "ずんだもん": "Ono_Anna", "れいむ": "Ono_Anna", "まりさ": "Ono_Anna"},
54 "qwen": {"四国めたん": "Ono_Anna", "ずんだもん": "Ono_Anna", "れいむ": "Ono_Anna", "まりさ": "Ono_Anna"},
55 "qwen_api": {"四国めたん": "Ono_Anna", "ずんだもん": "Ono_Anna", "れいむ": "Ono_Anna", "まりさ": "Ono_Anna"},
56 "qwen-api": {"四国めたん": "Ono_Anna", "ずんだもん": "Ono_Anna", "れいむ": "Ono_Anna", "まりさ": "Ono_Anna"},
57 "irodori": {"四国めたん": "default", "ずんだもん": "default", "れいむ": "default", "まりさ": "default"},
58 "irodori-tts": {"四国めたん": "default", "ずんだもん": "default", "れいむ": "default", "まりさ": "default"},
59 "irodori_api": {"四国めたん": "none", "ずんだもん": "none", "れいむ": "none", "まりさ": "none"},
60 "irodori-api": {"四国めたん": "none", "ずんだもん": "none", "れいむ": "none", "まりさ": "none"},
61 "aquestalkplayer": {"四国めたん": "れいむ", "ずんだもん": "まりさ", "れいむ": "れいむ", "まりさ": "まりさ", "青山龍星": "青山龍星"},
62 "openai": {"四国めたん": "nova", "ずんだもん": "shimmer", "れいむ": "alloy", "まりさ": "fable"},
63 "gemini": {"四国めたん": "Kore", "ずんだもん": "Puck", "れいむ": "Kore", "まりさ": "Charon"},
64 "edge": {"四国めたん": "Nanami", "ずんだもん": "Keita", "れいむ": "Nanami", "まりさ": "Keita"},
65 "edge_tts": {"四国めたん": "Nanami", "ずんだもん": "Keita", "れいむ": "Nanami", "まりさ": "Keita"},
66 "edge-tts": {"四国めたん": "Nanami", "ずんだもん": "Keita", "れいむ": "Nanami", "まりさ": "Keita"},
67}
68
69pause = 0
70
71
72def terminate():
73 if pause:
74 input("\nPress ENTER to terminate>>\n")
75 exit()
76
77def initialize():
78 parser = argparse.ArgumentParser(description="統合TTS (pyttsx3, WinRT, VOICEVOX, Qwen3-TTS, Irodori-TTS, AquesTalkPlayer, OpenAI, Gemini, Edge TTS) CLIツール")
79 parser.add_argument("--tts", "-t", choices=["pyttsx3", "winrt", "voicevox", "qwen3", "qwen", "qwen_api", "qwen-api", "irodori", "irodori-tts", "irodori_api", "irodori-api", "aquestalkplayer", "atp", "openai", "gemini", "edge", "edge_tts", "edge-tts"], default=DEFAULT_ENGINE, help="TTSエンジンを選択")
80 parser.add_argument("--endpoint", type=str, default=DEFAULT_VOICEVOX_ENDPOINT, help="VOICEVOX Engineのendpoint")
81# parser.add_argument("--language", default = 'japanese', help="pyttsx3 で使用する言語")
82 parser.add_argument("--monologue", "-m", type=int, default=0, help="独話形式 (カンマのない行も読み込む)")
83 parser.add_argument("--voices", "-v", type=str, default="", help="voice_map の上書き (key=val;key=val)")
84 parser.add_argument("--replace", "-r", type=str, default="", help="文字列置換ルール (key=val;key=val)")
85
86 parser.add_argument("--infile", "-i", type=str, default=DEFAULT_INPUT, help="入力元 ('clip' またはファイルパス)")
87 parser.add_argument("--outfile", "-o", type=str, default="", help="出力音声ファイル (未指定の場合、リアルタイム再生)")
88
89 parser.add_argument("--temp_dir", type=str, default=DEFAULT_TEMP_DIR, help="一時ファイルを作成するディレクトリ名 (ファイル出力型TTSで使用)")
90 parser.add_argument("--list", action="store_true", help="利用可能な voices を表示して終了")
91 parser.add_argument("--map", action="store_true", help="voice map を表示して終了")
92 parser.add_argument("--pause", "-p", type=int, default=0, help="終了時に入力待ちする")
93 parser.add_argument("--wait_for_clipboard", type=int, default=1, help="Clipbordから適すつを取得する際に入力待ちする")
94
95 parser.add_argument("--speak_rate", type=int, default=150, help="pyttsx3 の読み上げ速度 (Word Per Minute)")
96 parser.add_argument("--fspeak_rate", type=float, default=1.0, help="VOICEVOX の読み上げ速度比 (標準: 1.0)")
97 parser.add_argument("--fspeak_pitch", type=float, default=0.0, help="VOICEVOX の超えの高さ (標準: 0.0)")
98
99 parser.add_argument("--aquestalk_path", type=str, default=DEFAULT_AQUESTALK_PATH, help="AquesTalkPlayer.exe の実行パス (AquesTalkPlayer使用時)")
100 parser.add_argument("--tinterval", type=float, default=0.5, help="AquesTalkPlayer/OpenAIの音声ファイル間に挿入する無音区間の長さ(秒、デフォルト 0.5)")
101
102 # Qwen3-TTS
103 parser.add_argument("--qwen3_language", type=str, default="Japanese", help="Qwen3-TTS language (default: Japanese)")
104 parser.add_argument("--qwen3_model_id", type=str, default="Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice", help="Qwen3-TTS model ID")
105 parser.add_argument("--qwen3_device", type=str, default="auto", help="Qwen3-TTS device: auto/cuda:0/cpu")
106 parser.add_argument("--qwen3_dtype", type=str, default="auto", choices=["auto", "bfloat16", "bf16", "float16", "fp16", "float32", "fp32"], help="Qwen3-TTS dtype")
107 parser.add_argument("--qwen3_instruct", type=str, default="", help="Qwen3-TTS CustomVoiceへの追加指示")
108 parser.add_argument("--qwen_api_server", type=str, default=os.environ.get("QWEN_TTS_API", ""), help="Qwen TTS Flask API root URL")
109 parser.add_argument("--qwen_api_timeout", type=float, default=1800, help="Qwen TTS API timeout in seconds")
110
111 # Irodori-TTS
112 parser.add_argument("--irodori_caption", type=str, default="落ち着いた自然な声で、明瞭に読み上げる。", help="Irodori-TTS voice/style caption")
113 parser.add_argument("--irodori_ref_wav", type=str, default="", help="Irodori-TTS reference WAV")
114 parser.add_argument("--irodori_ref_wavs", type=str, default="", help="Irodori-TTS reference WAVs (;区切り)")
115 parser.add_argument("--irodori_model_id", type=str, default="Aratako/Irodori-TTS-v4.1-Small", help="Irodori-TTS model ID")
116 parser.add_argument("--irodori_device", type=str, default="auto", help="Irodori-TTS device: auto/cuda/cuda:0/cpu")
117 parser.add_argument("--irodori_precision", type=str, default="auto", choices=["auto", "bf16", "bfloat16", "fp32", "float32"], help="Irodori-TTS precision")
118 parser.add_argument("--irodori_codec_device", type=str, default=None, help="Irodori codec device (未指定ならmodel device)")
119 parser.add_argument("--irodori_codec_precision", type=str, default=None, help="Irodori codec precision (未指定ならmodel precision)")
120 parser.add_argument("--irodori_num_steps", type=int, default=40, help="Irodori-TTS sampling steps")
121 parser.add_argument("--irodori_cfg_scale_text", type=float, default=3.5, help="Irodori-TTS CFG scale for text")
122 parser.add_argument("--irodori_cfg_scale_caption", type=float, default=3.0, help="Irodori-TTS CFG scale for caption")
123 parser.add_argument("--irodori_cfg_scale_speaker", type=float, default=5.0, help="Irodori-TTS CFG scale for speaker/reference")
124 parser.add_argument("--irodori_duration_scale", type=float, default=1.0, help="Irodori-TTS duration scale")
125 parser.add_argument("--irodori_seed", type=str, default="0", help="Irodori-TTS seed; none/randomでランダム")
126 parser.add_argument("--irodori_lora_adapter", type=str, default="", help="Irodori-TTS LoRA adapter path")
127
128 # Irodori-TTS Flask API (GPU non-resident server)
129 parser.add_argument("--irodori_api_server", type=str, default=os.environ.get("IRODORI_TTS_API", "http://192.168.27.18:8002"), help="Irodori-TTS Flask API root URL")
130 parser.add_argument("--irodori_api_timeout", type=float, default=1800, help="Irodori-TTS API timeout in seconds")
131 parser.add_argument("--irodori_api_sway", action="store_true", help="Irodori APIでsway samplingを使用")
132 parser.add_argument("--irodori_api_sway_coeff", type=float, default=-1.0, help="Irodori API sway coefficient")
133
134 parser.add_argument("--instruction", type=str, default="", help="TTS APIへの追加指示 (OpenAI/Gemini使用時)")
135 parser.add_argument("--gemini_model", type=str, default="gemini-3.8-flash-tts", help="Gemini TTS model ID")
136 parser.add_argument("--gemini_api_key", type=str, default=os.environ.get("GEMINI_API_KEY", ""), help="Gemini API key (通常は GEMINI_API_KEY を使用)")
137 parser.add_argument("--edge_rate", type=str, default="+0%", help="Edge TTS speed (例: +20%%, -10%%)")
138 parser.add_argument("--edge_volume", type=str, default="+0%", help="Edge TTS volume (例: +10%%, -20%%)")
139 parser.add_argument("--edge_pitch", type=str, default="+0Hz", help="Edge TTS pitch (例: +5Hz, -10Hz)")
140
141 args = parser.parse_args()
142
143 # 独話でvoice指定が空の場合の既定値
144 if not args.voices:
145 if args.tts in ("qwen3", "qwen", "qwen_api", "qwen-api"):
146 args.voices = "Ono_Anna"
147 elif args.tts in ("irodori", "irodori-tts"):
148 args.voices = "default"
149 elif args.tts in ("irodori_api", "irodori-api"):
150 args.voices = "none"
151 elif args.tts == "gemini":
152 args.voices = "Kore"
153 elif args.tts in ("edge", "edge_tts", "edge-tts"):
154 args.voices = "Nanami"
155
156 # 空文字はバックエンド側では未指定として扱う
157 if args.irodori_ref_wav == "":
158 args.irodori_ref_wav = None
159 if args.irodori_ref_wavs == "":
160 args.irodori_ref_wavs = None
161 if args.irodori_lora_adapter == "":
162 args.irodori_lora_adapter = None
163 if args.qwen3_instruct == "":
164 args.qwen3_instruct = None
165
166 return args
167
168def main():
169 global pause
170
171 print()
172 print(f"\n===== 統合TTS CLIツール speak.py =====")
173
174 args = initialize()
175 pause = args.pause
176 if args.infile == "": args.infile = "clip"
177
178 print(f"TTS engine : {args.tts}")
179 print(f"is monologue: {args.monologue}")
180 print(f"Input : {args.infile}")
181 print(f"Output : {args.outfile}")
182# print(f"Language: {args.language}")
183 print(f"pyttsx3 speak_rate: {args.speak_rate}")
184 print(f"VOICEVOX speak_rate: {args.fspeak_rate}")
185 print(f"VOICEVOX speak_pitch: {args.fspeak_pitch}")
186 print(f"VOICEVOX Engine endpoint: {args.endpoint}")
187 print(f"wait_for_clipboard: {args.wait_for_clipboard}")
188
189 if args.tts in ("qwen3", "qwen"):
190 print(f"Qwen3 model : {args.qwen3_model_id}")
191 print(f"Qwen3 device: {args.qwen3_device}")
192 print(f"Qwen3 dtype : {args.qwen3_dtype}")
193 print(f"Qwen3 lang : {args.qwen3_language}")
194 print(f"Qwen3 voice : {args.voices}")
195 print(f"Qwen3 instruct: {args.qwen3_instruct}")
196
197 if args.tts in ("qwen_api", "qwen-api"):
198 print(f"Qwen API server: {args.qwen_api_server}")
199 print(f"Qwen API timeout: {args.qwen_api_timeout}")
200 print(f"Qwen model : {args.qwen3_model_id}")
201 print(f"Qwen language : {args.qwen3_language}")
202 print(f"Qwen voice : {args.voices}")
203
204 if args.tts in ("irodori", "irodori-tts"):
205 print(f"Irodori model : {args.irodori_model_id}")
206 print(f"Irodori device : {args.irodori_device}")
207 print(f"Irodori precision: {args.irodori_precision}")
208 print(f"Irodori caption : {args.irodori_caption}")
209 print(f"Irodori ref_wav : {args.irodori_ref_wav}")
210 print(f"Irodori ref_wavs : {args.irodori_ref_wavs}")
211 print(f"Irodori steps : {args.irodori_num_steps}")
212 print(f"Irodori duration : {args.irodori_duration_scale}")
213 print(f"Irodori seed : {args.irodori_seed}")
214
215 if args.tts in ("irodori_api", "irodori-api"):
216 print(f"Irodori API server : {args.irodori_api_server}")
217 print(f"Irodori API timeout : {args.irodori_api_timeout}")
218 print(f"Irodori caption : {args.irodori_caption}")
219 print(f"Irodori steps : {args.irodori_num_steps}")
220 print(f"Irodori duration : {args.irodori_duration_scale}")
221 print(f"Irodori API voice : {args.voices}")
222
223 if args.tts == "gemini":
224 print(f"Gemini model : {args.gemini_model}")
225 print(f"Gemini voice : {args.voices}")
226
227 if args.tts in ("edge", "edge_tts", "edge-tts"):
228 print(f"Edge voice : {args.voices}")
229 print(f"Edge rate : {args.edge_rate}")
230 print(f"Edge volume : {args.edge_volume}")
231 print(f"Edge pitch : {args.edge_pitch}")
232
233# endpoint, aquestalk_pathはargsで渡す
234 tktts = tkTTS(tts_name = args.tts, config = args)
235
236 if args.list:
237 print()
238 tktts.list_available_voices()
239 terminate()
240
241 if args.map:
242 print()
243 tktts.show_voice_map(args.infile, args.voices, VOICE_MAPS, args.monologue)
244 terminate()
245
246 print()
247 print(f"[{args.infile}]を解析します:")
248 dialogue = tktts.load_text(args.infile, args.monologue, wait_for_clipboard = args.wait_for_clipboard)
249 if not dialogue:
250 print("エラー: 有効なテキストデータが取得できませんでした。")
251 if not args.monologue:
252 print(" 対話形式でない場合は --monologue=1 オプションをつけてください。")
253 terminate()
254
255 speakers_in_file = tktts.get_speakers_from_dialogue(dialogue)
256 print(f" Speakers in [{args.infile}]")
257 for idx, sp in enumerate(speakers_in_file):
258 print(f" {idx:02d}: {sp}")
259
260 current_voice_map = tktts.update_voice_map(voice_map = VOICE_MAPS,
261 voices = args.voices, speakers = speakers_in_file)
262
263 print()
264 print("=== 置換辞書 ===")
265 replacements = tktts.parse_kv_string(args.replace)
266 if replacements:
267 for key, val in replacements.items():
268 print(f" {key}: {val}")
269 else:
270 print(" (なし)")
271
272 print()
273 print(f"Voice map updated:")
274 for key, val in current_voice_map.items():
275 if type(key) is str:
276 print(f" (speaker) {key}: (voice) {val}")
277 for key, val in current_voice_map.items():
278 if type(key) is not str and type(key) is not int:
279 print(f" (speaker) {key}: (voice) {val}")
280 for key, val in current_voice_map.items():
281 if type(key) is int:
282 print(f" (speaker) {key}: (voice) {val}")
283
284 print("=== 検出された話者とvoice ===")
285 print(f"Voice map updated;", current_voice_map)
286 for s in sorted(speakers_in_file):
287 if s is None or s == "":
288 voice = current_voice_map.get(s, None)
289 if voice is None: voice = current_voice_map.get(0, None)
290 print(f" (独話): {voice}")
291 else:
292 s = tktts.normalize_speaker(s, args.tts)
293 print(f" (speaker) {s}: (voice) {current_voice_map.get(s, '未設定')}")
294
295 print()
296 print("--- 読み上げ処理開始 ---")
297 ret = tktts.speak_dialogue(
298 config = args, dialogue = dialogue,
299 voice_map = current_voice_map, replacements = replacements)
300 if not ret:
301 print("❌ 読み上げ処理に失敗したため、出力ファイルは作成されませんでした。")
302 terminate()
303
304 print("--- 処理完了 ---")
305
306
307if __name__ == "__main__":
308 main()
309 terminate()