Add stream infer for Bert-VITS2.

Fix vits stream infer.

TTSManager.py

@@ -244,6 +244,18 @@ def vits_infer(self, state, encode=True):
         audio = np.concatenate(audios, axis=0)
         return self.encode(sampling_rate, audio, state["format"]) if encode else audio
 
+    def stream_vits_infer(self, task, fname=None):
+        format = task.get("format", "wav")
+        voice_obj = self._voice_obj[ModelType.VITS][task.get("id")][1]
+        task["id"] = self._voice_obj[ModelType.VITS][task.get("id")][0]
+        sampling_rate = voice_obj.sampling_rate
+        genertator = voice_obj.get_stream_audio(task, auto_break=True)
+        # audio = BytesIO()
+        for chunk in genertator:
+            encoded_audio = self.encode(sampling_rate, chunk, format)
+            for encoded_audio_chunk in self.generate_audio_chunks(encoded_audio):
+                yield encoded_audio_chunk
+
     def stream_vits_infer(self, state, fname=None):
         model = self.get_model(ModelType.VITS, state["id"])
         state["id"] = self.get_real_id(ModelType.VITS, state["id"])
@@ -266,11 +278,11 @@ def stream_vits_infer(self, state, fname=None):
             if i < sentences_num - 1:
                 audios.append(brk)
 
-        audio = np.concatenate(audios, axis=0)
-        encoded_audio = self.encode(sampling_rate, audio, state["format"])
+            audio = np.concatenate(audios, axis=0)
+            encoded_audio = self.encode(sampling_rate, audio, state["format"])
 
-        for encoded_audio_chunk in self.generate_audio_chunks(encoded_audio):
-            yield encoded_audio_chunk
+            for encoded_audio_chunk in self.generate_audio_chunks(encoded_audio):
+                yield encoded_audio_chunk
         #     if getattr(config, "SAVE_AUDIO", False):
         #         audio.write(encoded_audio.getvalue())
         # if getattr(config, "SAVE_AUDIO", False):
@@ -357,3 +369,28 @@ def bert_vits2_infer(self, state, encode=True):
         audio = np.concatenate(audios)
 
         return self.encode(sampling_rate, audio, state["format"]) if encode else audio
+
+    def stream_bert_vits2_infer(self, state, fname=None):
+        model = self.get_model(ModelType.BERT_VITS2, state["id"])
+        state["id"] = self.get_real_id(ModelType.BERT_VITS2, state["id"])
+
+        # 去除所有多余的空白字符
+        if state["text"] is not None:
+            state["text"] = re.sub(r'\s+', ' ', state["text"]).strip()
+        sampling_rate = model.sampling_rate
+
+        sentences_list = split_by_language(state["text"], state["speaker_lang"])
+
+        # audios = []
+
+        for (text, lang) in sentences_list:
+            sentences = sentence_split(text, state["max"])
+            for sentence in sentences:
+                audio = model.infer(sentence, state["id"], lang, state["sdp_ratio"], state["noise"],
+                                    state["noise"], state["length"])
+                # audios.append(audio)
+                # audio = np.concatenate(audios, axis=0)
+                encoded_audio = self.encode(sampling_rate, audio, state["format"])
+
+                for encoded_audio_chunk in self.generate_audio_chunks(encoded_audio):
+                    yield encoded_audio_chunk

tts_app/templates/pages/index.html

@@ -295,6 +295,11 @@ <h1 class="w-100">
                         <source src="" type="audio/mp3"/>
                         Your browser does not support the audio element.
                     </audio>
+                    <div class="mb-3 form-check">
+                        <input type="checkbox" id="streaming3" onchange="updateLink()">
+                        <label class="form-check-label" data-toggle="tooltip" data-placement="top"
+                               title="按照max分段推理文本，推理好一段即输出，无需等待所有文本都推理完毕">流式响应</label>
+                    </div>
                 </div>
             </div>
         </div>
@@ -392,6 +397,7 @@ <h1 class="w-100">
         } else if (model_type == 3) {
             var sdp_ratio = document.getElementById("input_sdp_ratio").value;
             var url = baseUrl + "/voice/bert-vits2?text=" + text + "&id=" + id;
+            var streaming = document.getElementById('streaming3');
         }
         if (format != "") {
             url += "&format=" + format;
@@ -411,7 +417,7 @@ <h1 class="w-100">
         if (max != "") {
             url += "&max=" + max;
         }
-        if (model_type == 1 && streaming.checked) {
+        if ((model_type == 1 || model_type == 3) && streaming.checked) {
             url += '&streaming=true';
         }
         if (model_type == 3 && sdp_ratio != "") {

tts_app/voice_api/views.py

@@ -387,6 +387,7 @@ def voice_bert_vits2_api():
         noisew = float(request_data.get("noisew", current_app.config.get("NOISEW", 0.8)))
         sdp_ratio = float(request_data.get("sdp_ratio", current_app.config.get("SDP_RATIO", 0.2)))
         max = int(request_data.get("max", current_app.config.get("MAX", 50)))
+        use_streaming = request_data.get('streaming', False, type=bool)
     except Exception as e:
         logger.error(f"[{ModelType.BERT_VITS2.value}] {e}")
         return make_response("parameter error", 400)
@@ -418,9 +419,13 @@ def voice_bert_vits2_api():
     if current_app.config.get("LANGUAGE_AUTOMATIC_DETECT", []) != []:
         speaker_lang = current_app.config.get("LANGUAGE_AUTOMATIC_DETECT")
 
+    if use_streaming and format.upper() != "MP3":
+        format = "mp3"
+        logger.warning("Streaming response only supports MP3 format.")
+
     fname = f"{str(uuid.uuid1())}.{format}"
     file_type = f"audio/{format}"
-    task = {"text": text,
+    state = {"text": text,
             "id": id,
             "format": format,
             "length": length,
@@ -430,11 +435,20 @@ def voice_bert_vits2_api():
             "max": max,
             "lang": lang,
             "speaker_lang": speaker_lang}
+
+    if use_streaming:
+        audio = tts_manager.stream_bert_vits2_infer(state)
+        response = make_response(audio)
+        response.headers['Content-Disposition'] = f'attachment; filename={fname}'
+        response.headers['Content-Type'] = file_type
+        return response
+    else:
+        t1 = time.time()
+        audio = tts_manager.bert_vits2_infer(state)
+        t2 = time.time()
+        logger.info(f"[{ModelType.BERT_VITS2.value}] finish in {(t2 - t1):.2f}s")
 
-    t1 = time.time()
-    audio = tts_manager.bert_vits2_infer(task)
-    t2 = time.time()
-    logger.info(f"[{ModelType.BERT_VITS2.value}] finish in {(t2 - t1):.2f}s")
+
 
     if current_app.config.get("SAVE_AUDIO", False):
         logger.debug(f"[{ModelType.BERT_VITS2.value}] {fname}")

utils/sentence.py

@@ -93,8 +93,8 @@ def sentence_split(text: str, max: int) -> list:
     if p < len(text):
         sentences_list.append(text[p:])
 
-    for i in sentences_list:
-        logging.debug(i)
+    # for i in sentences_list:
+    #     logging.debug(i)
 
     return sentences_list