fix: add seed

Signed-off-by: yihong0618 <zouzou0208@gmail.com>

.gitignore

@@ -168,4 +168,5 @@ nohup.out
 *.mp4
 *.pdf
 .pdm-python
-tts.wav
+tts.wav
+tts_pro.wav

handlers/tts.py

@@ -16,30 +16,49 @@ if USE_CHATTTS:
     chat.load_models()
     lock = threading.Lock()  # Initialize a lock
 
-    def generate_tts_wav(prompt):
+    def save_data_to_wav(filename, data):
-        texts = [
-            prompt,
-        ]
-        wavs = chat.infer(texts, use_decoder=True)
-        output_filename = "tts.wav"
-        audio_data = np.array(
-            wavs[0], dtype=np.float32
-        )  # Ensure the data type is correct
         sample_rate = 24000
-        # Normalize the audio data to 16-bit PCM range
-        audio_data = (audio_data * 32767).astype(np.int16)
-
         # Open a .wav file to write into
-        with wave.open(output_filename, "w") as wf:
+        with wave.open(filename, "w") as wf:
             wf.setnchannels(1)  # Mono channel
             wf.setsampwidth(2)  # 2 bytes per sample
             wf.setframerate(sample_rate)
-            wf.writeframes(audio_data.tobytes())
+            wf.writeframes(data.tobytes())
 
-        print(f"Audio has been saved to {output_filename}")
+    def generate_tts_wav(prompt, seed=None):
+        texts = [
+            prompt,
+        ]
+        if seed:
+            r = chat.sample_random_speaker(seed)
+            params_infer_code = {
+                "spk_emb": r,  # add sampled speaker
+                "temperature": 0.3,  # using custom temperature
+                "top_P": 0.7,  # top P decode
+                "top_K": 20,  # top K decode
+            }
+            wavs = chat.infer(
+                texts, use_decoder=True, params_infer_code=params_infer_code
+            )
+            output_filename = "tts_pro.wav"
+        else:
+            wavs = chat.infer(texts, use_decoder=True)
+            output_filename = "tts.wav"
+
+        audio_data = np.array(
+            wavs[0], dtype=np.float32
+        )  # Ensure the data type is correct
+        # Normalize the audio data to 16-bit PCM range
+        audio_data = (audio_data * 32767).astype(np.int16)
+        save_data_to_wav(output_filename, audio_data)
+
+        if seed:
+            print(f"Audio has been saved to {output_filename} with seed {seed}")
+        else:
+            print(f"Audio has been saved to {output_filename}")
 
     def tts_handler(message: Message, bot: TeleBot):
-        """pretty tts: /tts <address>"""
+        """pretty tts: /tts <prompt>"""
         bot.reply_to(
             message, f"Generating ChatTTS may take some time please wait some time."
         )
@@ -59,6 +78,37 @@ if USE_CHATTTS:
             print(e)
             bot.reply_to(message, "tts error")
 
+    def tts_pro_handler(message: Message, bot: TeleBot):
+        """pretty tts_pro: /tts_pro <seed>,<prompt>"""
+        m = message.text.strip()
+        prompt = m.strip()
+        seed = prompt.split(",")[0]
+        bot.reply_to(
+            message,
+            f"Generating ChatTTS with seed: {seed} may take some time please wait some time.",
+        )
+        if not seed.isdigit():
+            bot.reply_to(message, "first argument must be a number")
+            return
+        prompt = prompt[len(str(seed)) + 1 :]
+        if len(prompt) > 150:
+            bot.reply_to(message, "prompt too long must length < 150")
+            return
+        try:
+            with lock:
+                generate_tts_wav(prompt, seed)
+            with open(f"tts_pro.wav", "rb") as audio:
+                bot.send_audio(
+                    message.chat.id, audio, reply_to_message_id=message.message_id
+                )
+        except Exception as e:
+            print(e)
+            bot.reply_to(message, "tts error")
+
     def register(bot: TeleBot) -> None:
         bot.register_message_handler(tts_handler, commands=["tts"], pass_bot=True)
         bot.register_message_handler(tts_handler, regexp="^tts:", pass_bot=True)
+        bot.register_message_handler(
+            tts_pro_handler, commands=["tts_pro"], pass_bot=True
+        )
+        bot.register_message_handler(tts_pro_handler, regexp="^tts_pro:", pass_bot=True)