Update and Fix

Change the frontend audio playback to use the POST method.
Disable CSRF protection for the voice_api blueprint.
Add an audio download button.
Fix cache clearing.
Resolve the npy loading issue in w2v2.

ModelManager.py

@@ -174,7 +174,7 @@ class ModelManager(Subject):
 
         if model_type == ModelType.W2V2_VITS:
             if self.emotion_reference is None:
-                self.emotion_reference = self.load_npy(config["model_config"]["dimensional_emotion_model"])
+                self.emotion_reference = self.load_npy(config["model_config"]["dimensional_emotion_npy"])
             model_args.update({"emotion_reference": self.emotion_reference,
                                "dimensional_emotion_model": self.dimensional_emotion_model})
 

TTSManager.py

@@ -130,7 +130,7 @@ class TTSManager(Observer):
                 length = float(element.attrib.get("length", root.attrib.get("length", config.LENGTH)))
                 noise = float(element.attrib.get("noise", root.attrib.get("noise", config.NOISE)))
                 noisew = float(element.attrib.get("noisew", root.attrib.get("noisew", config.NOISEW)))
-                max = int(element.attrib.get("max", root.attrib.get("max", "0")))
+                segment_size = int(element.attrib.get("segment_size", root.attrib.get("segment_size", "0")))
                 # 不填写默认就是vits
                 model_type = element.attrib.get("model_type", root.attrib.get("model_type", "vits"))
                 # w2v2-vits/emotion-vits才有emotion
@@ -170,7 +170,7 @@ class TTSManager(Observer):
                                             "length": length,
                                             "noise": noise,
                                             "noisew": noisew,
-                                            "max": max,
+                                            "segment_size": segment_size,
                                             "model_type": model_type,
                                             "emotion": emotion,
                                             "sdp_ratio": sdp_ratio,
@@ -228,7 +228,7 @@ class TTSManager(Observer):
             state["text"] = re.sub(r'\s+', ' ', state["text"]).strip()
         sampling_rate = model.sampling_rate
 
-        sentences_list = sentence_split_and_markup(state["text"], state["max"], state["lang"], state["speaker_lang"])
+        sentences_list = sentence_split_and_markup(state["text"], state["segment_size"], state["lang"], state["speaker_lang"])
         # 停顿0.5s，避免语音分段合成再拼接后的连接突兀
         brk = np.zeros(int(0.5 * sampling_rate), dtype=np.int16)
 
@@ -265,7 +265,7 @@ class TTSManager(Observer):
             state["text"] = re.sub(r'\s+', ' ', state["text"]).strip()
         sampling_rate = model.sampling_rate
 
-        sentences_list = sentence_split_and_markup(state["text"], state["max"], state["lang"], state["speaker_lang"])
+        sentences_list = sentence_split_and_markup(state["text"], state["segment_size"], state["lang"], state["speaker_lang"])
         # 停顿0.5s，避免语音分段合成再拼接后的连接突兀
         brk = np.zeros(int(0.5 * sampling_rate), dtype=np.int16)
 
@@ -307,7 +307,7 @@ class TTSManager(Observer):
 
         sampling_rate = model.sampling_rate
 
-        sentences_list = sentence_split_and_markup(state["text"], state["max"], state["lang"], state["speaker_lang"])
+        sentences_list = sentence_split_and_markup(state["text"], state["segment_size"], state["lang"], state["speaker_lang"])
         # 停顿0.5s，避免语音分段合成再拼接后的连接突兀
         brk = np.zeros(int(0.5 * sampling_rate), dtype=np.int16)
 
@@ -361,7 +361,7 @@ class TTSManager(Observer):
         audios = []
 
         for (text, lang) in sentences_list:
-            sentences = sentence_split(text, state["max"])
+            sentences = sentence_split(text, state["segment_size"])
             for sentence in sentences:
                 audio = model.infer(sentence, state["id"], lang, state["sdp_ratio"], state["noise"],
                                     state["noise"], state["length"])
@@ -384,7 +384,7 @@ class TTSManager(Observer):
         # audios = []
 
         for (text, lang) in sentences_list:
-            sentences = sentence_split(text, state["max"])
+            sentences = sentence_split(text, state["segment_size"])
             for sentence in sentences:
                 audio = model.infer(sentence, state["id"], lang, state["sdp_ratio"], state["noise"],
                                     state["noise"], state["length"])

app.py

@@ -5,6 +5,7 @@ from flask_apscheduler import APScheduler
 from flask_login import LoginManager
 from flask_wtf import CSRFProtect
 
+from utils.data_utils import clean_folder
 from utils.phrases_dict import phrases_dict_init
 from tts_app import frontend, voice_api, auth, admin
 from utils.config_manager import global_config
@@ -17,13 +18,15 @@ app.config.update(global_config)
 
 phrases_dict_init()
 
+csrf = CSRFProtect(app)
+# 禁用tts api请求的CSRF防护
+csrf.exempt(voice_api)
+
 if app.config.get("IS_ADMIN_ENABLED", False):
     login_manager = LoginManager()
     login_manager.init_app(app)
     login_manager.login_view = 'auth.login'
 
-    csrf = CSRFProtect(app)
-
 
     @login_manager.user_loader
     def load_user(user_id):
@@ -44,6 +47,13 @@ app.register_blueprint(voice_api, url_prefix='/voice')
 if app.config.get("IS_ADMIN_ENABLED", False):
     app.register_blueprint(auth, url_prefix=app.config.get("ADMIN_ROUTE", "/admin"))
     app.register_blueprint(admin, url_prefix=app.config.get("ADMIN_ROUTE", "/admin"))
+    
+# regular cleaning
+@scheduler.task('interval', id='clean_task', seconds=app.config.get("CLEAN_INTERVAL_SECONDS", 3600),
+                misfire_grace_time=900)
+def clean_task():
+    clean_folder(app.config["UPLOAD_FOLDER"])
+    clean_folder(app.config["CACHE_PATH"])
 
 if __name__ == '__main__':
     app.run(host='0.0.0.0', port=app.config.get("PORT", 23456), debug=app.config.get("DEBUG", False))

config.py

@@ -109,8 +109,8 @@ NOISE = 0.33
 NOISEW = 0.4
 
 # 长文本分段阈值，max<=0表示不分段.
-# Batch processing threshold. Text will not be processed in batches if max<=0
-MAX = 50
+# Batch processing threshold. Text will not be processed in batches if segment<=0
+SEGMENT = 50
 
 # Bert_VITS2
 SDP_RATIO = 0.2

tts_app/static/css/pages/index.css

@@ -67,3 +67,8 @@
 label {
     margin-bottom: 0.5rem;
 }
+
+button {
+    margin-left: 10px;
+    margin-right: 10px;
+}

tts_app/static/js/index.js

@@ -0,0 +1,285 @@
+var baseUrl = getProtocol() + getUrl();
+var currentModelPage = 1;
+var vitsSpeakersCount = 0;
+var w2v2SpeakersCount = 0;
+var bertVits2SpeakersCount = 0;
+
+function speakersInit() {
+    $.ajax({
+        url: '/voice/speakers',
+        type: 'GET',
+        dataType: 'json',
+        success: function (data) {
+            vitsSpeakersCount = data['VITS'].length;
+            w2v2SpeakersCount = data['W2V2-VITS'].length;
+            bertVits2SpeakersCount = data['BERT-VITS2'].length;
+            showModelContentBasedOnStatus();
+        },
+        error: function (xhr, status, error) {
+            console.error('Request failed with status', status, 'and error', error);
+        }
+    });
+}
+
+$(function () {
+    $('[data-toggle="tooltip"]').tooltip()
+})
+
+function getProtocol() {
+    return 'https:' == location.protocol ? "https://" : "http://";
+}
+
+function getUrl() {
+    let url = window.location.host;
+    return url;
+}
+
+
+function setBaseUrl() {
+    let text = document.getElementById("input_text" + currentModelPage).value;
+    let id = document.getElementById("input_id" + currentModelPage).value;
+
+    let vits_link = document.getElementById("vits_link");
+    let speakers_link = document.getElementById("speakers_link");
+
+    let vits_url = baseUrl + "/voice/vits?text=" + text + "&id=" + id;
+    let speakers_url = baseUrl + "/voice/speakers";
+
+    vits_link.href = vits_url;
+    vits_link.textContent = vits_url;
+
+    speakers_link.href = speakers_url;
+    speakers_link.textContent = speakers_url;
+}
+
+function getLink() {
+    let text = document.getElementById("input_text" + currentModelPage).value;
+    let id = document.getElementById("input_id" + currentModelPage).value;
+    let format = document.getElementById("input_format" + currentModelPage).value;
+    let lang = document.getElementById("input_lang" + currentModelPage).value;
+    let length = document.getElementById("input_length" + currentModelPage).value;
+    let noise = document.getElementById("input_noise" + currentModelPage).value;
+    let noisew = document.getElementById("input_noisew" + currentModelPage).value;
+    let max = document.getElementById("input_max" + currentModelPage).value;
+
+    let url = baseUrl
+    let streaming = null;
+    let sdp_ratio = null;
+    if (currentModelPage == 1) {
+        streaming = document.getElementById('streaming1');
+        url += "/voice/vits?text=" + text + "&id=" + id;
+    } else if (currentModelPage == 2) {
+        let emotion = document.getElementById('emotion').value;
+        url += "/voice/w2v2-vits?text=" + text + "&id=" + id + "&emotion=" + emotion;
+    } else if (currentModelPage == 3) {
+        sdp_ratio = document.getElementById("input_sdp_ratio").value;
+        streaming = document.getElementById('streaming3');
+        url += "/voice/bert-vits2?text=" + text + "&id=" + id;
+
+    } else {
+        console.error("Invalid model page: ", currentModelPage);
+        return null;
+    }
+    if (format != "") {
+        url += "&format=" + format;
+    }
+    if (lang != "") {
+        url += "&lang=" + lang;
+    }
+    if (length != "") {
+        url += "&length=" + length;
+    }
+    if (noise != "") {
+        url += "&noise=" + noise;
+    }
+    if (noisew != "") {
+        url += "&noisew=" + noisew;
+    }
+    if (max != "") {
+        url += "&max=" + max;
+    }
+
+    if (currentModelPage == 1) {
+        if (streaming.checked)
+            url += '&streaming=true';
+    } else if (currentModelPage == 3) {
+        if (streaming.checked)
+            url += '&streaming=true';
+        if (sdp_ratio != "")
+            url += "&sdp_ratio=" + sdp_ratio;
+    }
+
+    return url;
+}
+
+function updateLink() {
+    let url = getLink();
+    let link = document.getElementById("vits_link");
+    link.href = url;
+    link.textContent = url;
+}
+
+function setAudioSourceByGet() {
+    if (currentModelPage == 1 && vitsSpeakersCount <= 0) {
+        alert("未加载VITS模型");
+        return;
+    }
+    if (currentModelPage == 2 && w2v2SpeakersCount <= 0) {
+        alert("未加载W2V2-VITS模型");
+        return;
+    }
+    if (currentModelPage == 3 && bertVits2SpeakersCount <= 0) {
+        alert("未加载Bert-VITS2模型");
+        return;
+    }
+    let url = getLink();
+
+    // Add a timestamp parameter to prevent browser caching
+    let timestamp = new Date().getTime();
+    url += '&t=' + timestamp;
+
+    let audioPlayer = document.getElementById("audioPlayer" + currentModelPage);
+    audioPlayer.src = url;
+    audioPlayer.play();
+}
+
+function setAudioSourceByPost() {
+    if (currentModelPage == 1 && vitsSpeakersCount <= 0) {
+        alert("未加载VITS模型");
+        return;
+    }
+    if (currentModelPage == 2 && w2v2SpeakersCount <= 0) {
+        alert("未加载W2V2-VITS模型");
+        return;
+    }
+    if (currentModelPage == 3 && bertVits2SpeakersCount <= 0) {
+        alert("未加载Bert-VITS2模型");
+        return;
+    }
+    let text = $("#input_text" + currentModelPage).val();
+    let id = $("#input_id" + currentModelPage).val();
+    let format = $("#input_format" + currentModelPage).val();
+    let lang = $("#input_lang" + currentModelPage).val();
+    let length = $("#input_length" + currentModelPage).val();
+    let noise = $("#input_noise" + currentModelPage).val();
+    let noisew = $("#input_noisew" + currentModelPage).val();
+    let max = $("#input_max" + currentModelPage).val();
+
+    let url = "";
+    let streaming = null;
+    let sdp_ratio = null;
+
+    if (currentModelPage == 1) {
+        url = baseUrl + "/voice/vits";
+        streaming = $("#streaming1")[0];
+    } else if (currentModelPage == 2) {
+        let emotion = $("#emotion").val();
+        url = baseUrl + "/voice/w2v2-vits";
+    } else if (currentModelPage == 3) {
+        sdp_ratio = $("#input_sdp_ratio").val();
+        url = baseUrl + "/voice/bert-vits2";
+        streaming = $("#streaming3")[0];
+    }
+
+    let data = {
+        text: text,
+        id: id,
+        format: format,
+        lang: lang,
+        length: length,
+        noise: noise,
+        noisew: noisew,
+        max: max
+    };
+
+    if ((currentModelPage == 1 || currentModelPage == 3) && streaming.checked) {
+        data.streaming = true;
+    }
+    if (currentModelPage == 3 && sdp_ratio != "") {
+
+        data.sdp_ratio = sdp_ratio;
+    }
+
+    $.ajax({
+        url: url,
+        method: 'POST',
+        data: JSON.stringify(data),
+        contentType: 'application/json',
+        responseType: 'blob',
+        xhrFields: {
+            responseType: 'blob'
+        },
+        success: function (response, status, xhr) {
+
+            let blob = new Blob([response], {type: 'audio/wav'});
+            let audioPlayer = document.getElementById("audioPlayer" + currentModelPage);
+            let audioFileName = getFileNameFromResponseHeader(xhr);
+            audioPlayer.setAttribute('data-file-name', audioFileName);
+            audioPlayer.src = URL.createObjectURL(blob);
+            audioPlayer.load();
+            audioPlayer.play();
+        },
+        error: function (error) {
+            console.error('Error:', error);
+            alert("无法获取音频数据");
+        }
+    });
+
+}
+
+function getFileNameFromResponseHeader(xhr) {
+    var contentDispositionHeader = xhr.getResponseHeader('Content-Disposition');
+    var matches = contentDispositionHeader.match(/filename=(.+)$/);
+    return matches ? matches[1] : 'audio.wav';  // 如果无法从响应头获取文件名，则使用默认值
+}
+
+function downloadAudio() {
+    let audioPlayer = document.getElementById("audioPlayer" + currentModelPage);
+    let audioFileName = audioPlayer.getAttribute('data-file-name') || 'audio.wav';
+
+    let downloadLink = document.createElement('a');
+    downloadLink.href = audioPlayer.src;
+
+    downloadLink.download = audioFileName;
+
+    document.body.appendChild(downloadLink);
+    downloadLink.click();
+    document.body.removeChild(downloadLink);
+}
+
+
+function showContent(index) {
+    const panes = document.querySelectorAll(".content-pane");
+    const buttons = document.querySelectorAll(".tab-button");
+    currentModelPage = index + 1;
+
+    for (let i = 0; i < panes.length; i++) {
+        if (i === index) {
+            panes[i].classList.add("active");
+            buttons[i].classList.add("active");
+
+        } else {
+            panes[i].classList.remove("active");
+            buttons[i].classList.remove("active");
+        }
+    }
+    updateLink();
+}
+
+function showModelContentBasedOnStatus() {
+    if (vitsSpeakersCount > 0) {
+        showContent(0);
+    } else if (w2v2SpeakersCount > 0) {
+        showContent(1);
+    } else if (bertVits2SpeakersCount > 0) {
+        showContent(2);
+    } else {
+        showContent(0);
+    }
+}
+
+$(document).ready(function () {
+    speakersInit();
+
+    setBaseUrl();
+});

tts_app/templates/pages/index.html

@@ -98,7 +98,9 @@
 
                 <div class="flex flex-wrap  w-100"
                      style="justify-content: center; align-items: center; height: 80px; margin-top: 20px; margin-bottom: 20px; border: 1px solid rgba(0,0,0,.125); border-radius: 0.25rem;">
-                    <button type="button" class="btn btn-outline-secondary" onclick="setAudioSource()"
+
+
+                    <button type="button" class="btn btn-outline-secondary" onclick="setAudioSourceByPost()"
                             style="margin-right: 10px">
                         播放器生成
                     </button>
@@ -106,6 +108,10 @@
                         <source src="" type="audio/mp3"/>
                         Your browser does not support the audio element.
                     </audio>
+                    <button type="button" class="btn btn-outline-secondary" onclick="downloadAudio()"
+                            style="margin-right: 10px">
+                        下载音频
+                    </button>
                     <div class="mb-3 form-check">
                         <input type="checkbox" id="streaming1" onchange="updateLink()">
                         <label class="form-check-label" data-toggle="tooltip" data-placement="top"
@@ -196,14 +202,19 @@
 
                 <div class="flex flex-wrap  w-100"
                      style="justify-content: center; align-items: center; height: 80px; margin-top: 20px; margin-bottom: 20px; border: 1px solid rgba(0,0,0,.125); border-radius: 0.25rem;">
-                    <button type="button" class="btn btn-outline-secondary" onclick="setAudioSource()"
-                            style="margin-right: 10px">
-                        播放器生成
-                    </button>
+
                     <audio id="audioPlayer2" controls>
                         <source src="" type="audio/mp3"/>
                         Your browser does not support the audio element.
                     </audio>
+                    <button type="button" class="btn btn-outline-secondary" onclick="setAudioSourceByPost()"
+                            style="margin-right: 10px">
+                        播放器生成
+                    </button>
+                    <button type="button" class="btn btn-outline-secondary" onclick="downloadAudio()"
+                            style="margin-right: 10px">
+                        下载音频
+                    </button>
                 </div>
             </div>
             <div class="content-pane w-100 flex-wrap" id="bert-vits2-page">
@@ -288,7 +299,7 @@
 
                 <div class="flex flex-wrap  w-100"
                      style="justify-content: center; align-items: center; height: 80px; margin-top: 20px; margin-bottom: 20px; border: 1px solid rgba(0,0,0,.125); border-radius: 0.25rem;">
-                    <button type="button" class="btn btn-outline-secondary" onclick="setAudioSource()"
+                    <button type="button" class="btn btn-outline-secondary" onclick="setAudioSourceByPost()"
                             style="margin-right: 10px">
                         播放器生成
                     </button>
@@ -296,6 +307,10 @@
                         <source src="" type="audio/mp3"/>
                         Your browser does not support the audio element.
                     </audio>
+                    <button type="button" class="btn btn-outline-secondary" onclick="downloadAudio()"
+                            style="margin-right: 10px">
+                        下载音频
+                    </button>
                     <div class="mb-3 form-check">
                         <input type="checkbox" id="streaming3" onchange="updateLink()">
                         <label class="form-check-label" data-toggle="tooltip" data-placement="top"
@@ -331,185 +346,6 @@
 
 {% include 'includes/scripts.html' %}
 
-<script>
-    $(function () {
-        $('[data-toggle="tooltip"]').tooltip()
-    })
-
-    function getProtocol() {
-        return 'https:' == location.protocol ? "https://" : "http://";
-    }
-
-    function getUrl() {
-        var url = window.location.host;
-        return url;
-    }
-
-    var baseUrl = getProtocol() + getUrl();
-    var model_type = 1;
-    var vits_status = false;
-    var w2v2_status = false;
-    var bert_vits2_status = false;
-    {% if vits_speakers_count > 0 %}
-        vits_status = true;
-    {% endif %}
-    {% if w2v2_speakers_count > 0 %}
-        w2v2_status = true;
-    {% endif %}
-    {% if bert_vits2_speakers_count > 0 %}
-        bert_vits2_status = true;
-    {% endif %}
-
-    $(document).ready(function () {
-        if (vits_status) {
-            $('#vits-button').addClass("active");
-            $('#vits-page').addClass("active");
-            model_type =  1;
-        } else if (w2v2_status) {
-            $('#w2v2-vits-button').addClass("active");
-            $('#w2v2-vits-page').addClass("active");
-            model_type =  2;
-        } else if (bert_vits2_status) {
-            $('#bert-vits2-button').addClass("active");
-            $('#bert-vits2-page').addClass("active");
-            model_type =  3;
-        } else {
-            $('#vits-button').addClass("active");
-            $('#vits-page').addClass("active");
-            model_type =  1;
-        }
-        setBaseUrl();
-    });
-
-
-    function setBaseUrl() {
-        var text = document.getElementById("input_text" + model_type).value;
-        var id = document.getElementById("input_id" + model_type).value;
-
-        var vits_link = document.getElementById("vits_link");
-        var speakers_link = document.getElementById("speakers_link");
-
-        var vits_url = baseUrl + "/voice/vits?text=" + text + "&id=" + id;
-        var speakers_url = baseUrl + "/voice/speakers";
-
-        vits_link.href = vits_url;
-        vits_link.textContent = vits_url;
-
-        speakers_link.href = speakers_url;
-        speakers_link.textContent = speakers_url;
-    }
-
-    function getLink() {
-        var text = document.getElementById("input_text" + model_type).value;
-        var id = document.getElementById("input_id" + model_type).value;
-        var format = document.getElementById("input_format" + model_type).value;
-        var lang = document.getElementById("input_lang" + model_type).value;
-        var length = document.getElementById("input_length" + model_type).value;
-        var noise = document.getElementById("input_noise" + model_type).value;
-        var noisew = document.getElementById("input_noisew" + model_type).value;
-        var max = document.getElementById("input_max" + model_type).value;
-
-        if (model_type == 1) {
-            var url = baseUrl + "/voice/vits?text=" + text + "&id=" + id;
-            var streaming = document.getElementById('streaming' + model_type);
-        } else if (model_type == 2) {
-            var emotion = document.getElementById('emotion').value;
-            var url = baseUrl + "/voice/w2v2-vits?text=" + text + "&id=" + id + "&emotion=" + emotion;
-        } 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;
-        }
-        if (lang != "") {
-            url += "&lang=" + lang;
-        }
-        if (length != "") {
-            url += "&length=" + length;
-        }
-        if (noise != "") {
-            url += "&noise=" + noise;
-        }
-        if (noisew != "") {
-            url += "&noisew=" + noisew;
-        }
-        if (max != "") {
-            url += "&max=" + max;
-        }
-        if ((model_type == 1 || model_type == 3) && streaming.checked) {
-            url += '&streaming=true';
-        }
-        if (model_type == 3 && sdp_ratio != "") {
-            url += "&sdp_ratio=" + sdp_ratio;
-        }
-
-        return url;
-    }
-
-    function updateLink() {
-        var url = getLink();
-        var link = document.getElementById("vits_link");
-        link.href = url;
-        link.textContent = url;
-    }
-
-    function setAudioSource() {
-        if (model_type == 1 && !vits_status) {
-            alert("未加载VITS模型");
-            return;
-        }
-        if (model_type == 2 && !w2v2_status) {
-            alert("未加载W2V2-VITS模型");
-            return;
-        }
-        if (model_type == 3 && !bert_vits2_status) {
-            alert("未加载Bert-VITS2模型");
-            return;
-        }
-        var url = getLink();
-
-        // Add a timestamp parameter to prevent browser caching
-        var timestamp = new Date().getTime();
-        url += '&t=' + timestamp;
-
-        var audioPlayer = document.getElementById("audioPlayer" + model_type);
-        audioPlayer.src = url;
-        audioPlayer.play();
-    }
-
-    function showContent(index) {
-        const panes = document.querySelectorAll(".content-pane");
-        const buttons = document.querySelectorAll(".tab-button");
-        model_type = index + 1;
-
-        for (let i = 0; i < panes.length; i++) {
-            if (i === index) {
-                panes[i].classList.add("active");
-                buttons[i].classList.add("active");
-
-            } else {
-                panes[i].classList.remove("active");
-                buttons[i].classList.remove("active");
-            }
-        }
-        updateLink();
-    }
-
-    document.querySelectorAll('.slider-group').forEach(function (group) {
-        group.addEventListener("input", function (event) {
-            if (event.target.matches('.slider')) {
-                let value = event.target.value;
-                group.querySelector('.slider-input').value = value;
-                group.querySelector('.slider-value').textContent = value;
-            } else if (event.target.matches('.slider-input')) {
-                let value = event.target.value;
-                group.querySelector('.slider').value = value;
-                group.querySelector('.slider-value').textContent = value;
-            }
-        });
-    });
-</script>
+<script src="{{ url_for('static', filename='js/index.js') }}"></script>
 </body>
 </html>

tts_app/voice_api/utils/__init__.py

@@ -1,14 +1,3 @@
-import os
-
-
-def clean_folder(folder_path):
-    for filename in os.listdir(folder_path):
-        file_path = os.path.join(folder_path, filename)
-        # 如果是文件，则删除文件。如果是文件夹则跳过。
-        if os.path.isfile(file_path):
-            os.remove(file_path)
-
-
 def check_is_none(item) -> bool:
     # none -> True, not none -> False
     return item is None or (isinstance(item, str) and str(item).isspace()) or str(item) == ""
@@ -16,4 +5,4 @@ def check_is_none(item) -> bool:
 
 def save_audio(audio, path):
     with open(path, "wb") as f:
-        f.write(audio)
+        f.write(audio)

tts_app/voice_api/views.py

@@ -1,3 +1,4 @@
+import os
 import time
 import uuid
 from io import BytesIO
@@ -15,6 +16,26 @@ from utils.data_utils import check_is_none
 voice_api = Blueprint("voice_api", __name__)
 
 
+def get_param(request_data, key, default, data_type=None):
+    if key == "segment_size" and "max" in request_data:
+        logger.warning(
+            "The 'max' parameter is deprecated and will be phased out in the future. Please use 'segment_size' instead.")
+        return get_param(request_data, "max", default, data_type)
+
+    value = request_data.get(key, "")
+    
+    if data_type:
+        try:
+            value = data_type(value)
+        except:
+            value = default
+
+    if value == "":
+        value = default
+
+    return value
+
+
 @voice_api.route('/speakers', methods=["GET", "POST"])
 def voice_speakers_api():
     return jsonify(model_manager.voice_speakers)
@@ -34,15 +55,15 @@ def voice_vits_api():
             else:
                 request_data = request.form
 
-        text = request_data.get("text", "")
-        id = int(request_data.get("id", current_app.config.get("ID", 0)))
-        format = request_data.get("format", current_app.config.get("FORMAT", "wav"))
-        lang = request_data.get("lang", current_app.config.get("LANG", "auto")).lower()
-        length = float(request_data.get("length", current_app.config.get("LENGTH", 1)))
-        noise = float(request_data.get("noise", current_app.config.get("NOISE", 0.667)))
-        noisew = float(request_data.get("noisew", current_app.config.get("NOISEW", 0.8)))
-        max = int(request_data.get("max", current_app.config.get("MAX", 50)))
-        use_streaming = request_data.get('streaming', False, type=bool)
+        text = get_param(request_data, "text", "", str)
+        id = get_param(request_data, "id", current_app.config.get("ID", 0), int)
+        format = get_param(request_data, "format", current_app.config.get("FORMAT", "wav"), str)
+        lang = get_param(request_data, "lang", current_app.config.get("LANG", "auto"), str).lower()
+        length = get_param(request_data, "length", current_app.config.get("LENGTH", 1), float)
+        noise = get_param(request_data, "noise", current_app.config.get("NOISE", 0.667), float)
+        noisew = get_param(request_data, "noisew", current_app.config.get("NOISEW", 0.8), float)
+        segment_size = get_param(request_data, "segment_size", current_app.config.get("SEGMENT_SIZE", 50), int)
+        use_streaming = get_param(request_data, 'streaming', False, bool)
     except Exception as e:
         logger.error(f"[{ModelType.VITS.value}] {e}")
         return make_response("parameter error", 400)
@@ -86,7 +107,7 @@ def voice_vits_api():
              "length": length,
              "noise": noise,
              "noisew": noisew,
-             "max": max,
+             "segment_size": segment_size,
              "lang": lang,
              "speaker_lang": speaker_lang}
 
@@ -116,12 +137,12 @@ def voice_hubert_api():
     if request.method == "POST":
         try:
             voice = request.files['upload']
-            id = int(request.form.get("id"))
-            format = request.form.get("format", current_app.config.get("LANG", "auto"))
-            length = float(request.form.get("length", current_app.config.get("LENGTH", 1)))
-            noise = float(request.form.get("noise", current_app.config.get("NOISE", 0.667)))
-            noisew = float(request.form.get("noisew", current_app.config.get("NOISEW", 0.8)))
-            use_streaming = request.form.get('streaming', False, type=bool)
+            id = get_param(request.form, "id", 0, int)
+            format = get_param(request.form, "format", current_app.config.get("LANG", "auto"))
+            length = get_param(request.form, "length", current_app.config.get("LENGTH", 1), float)
+            noise = get_param(request.form, "noise", current_app.config.get("NOISE", 0.667), float)
+            noisew = get_param(request.form, "noisew", current_app.config.get("NOISEW", 0.8), float)
+            use_streaming = get_param(request.form, 'streaming', False, bool)
         except Exception as e:
             logger.error(f"[{ModelType.HUBERT_VITS.value}] {e}")
             return make_response("parameter error", 400)
@@ -181,17 +202,17 @@ def voice_w2v2_api():
             else:
                 request_data = request.form
 
-        text = request_data.get("text", "")
-        id = int(request_data.get("id", current_app.config.get("ID", 0)))
-        format = request_data.get("format", current_app.config.get("FORMAT", "wav"))
-        lang = request_data.get("lang", current_app.config.get("LANG", "auto")).lower()
-        length = float(request_data.get("length", current_app.config.get("LENGTH", 1)))
-        noise = float(request_data.get("noise", current_app.config.get("NOISE", 0.667)))
-        noisew = float(request_data.get("noisew", current_app.config.get("NOISEW", 0.8)))
-        max = int(request_data.get("max", current_app.config.get("MAX", 50)))
-        emotion = int(request_data.get("emotion", current_app.config.get("EMOTION", 0)))
-        emotion_reference = request_data.get("emotion_reference", None)
-        use_streaming = request_data.get('streaming', False, type=bool)
+        text = get_param(request_data, "text", "", str)
+        id = get_param(request_data, "id", current_app.config.get("ID", 0), int)
+        format = get_param(request_data, "format", current_app.config.get("FORMAT", "wav"), str)
+        lang = get_param(request_data, "lang", current_app.config.get("LANG", "auto"), str).lower()
+        length = get_param(request_data, "length", current_app.config.get("LENGTH", 1), float)
+        noise = get_param(request_data, "noise", current_app.config.get("NOISE", 0.667), float)
+        noisew = get_param(request_data, "noisew", current_app.config.get("NOISEW", 0.8), float)
+        segment_size = get_param(request_data, "segment_size", current_app.config.get("SEGMENT_SIZE", 50), int)
+        emotion = get_param(request_data, "emotion", current_app.config.get("EMOTION", 0), int)
+        emotion_reference = get_param(request_data, "emotion_reference", None, str)
+        use_streaming = get_param(request_data, 'streaming', False, bool)
     except Exception as e:
         logger.error(f"[{ModelType.W2V2_VITS.value}] {e}")
         return make_response(f"parameter error", 400)
@@ -235,7 +256,7 @@ def voice_w2v2_api():
             "length": length,
             "noise": noise,
             "noisew": noisew,
-            "max": max,
+            "segment_size": segment_size,
             "lang": lang,
             "emotion": emotion,
             "emotion_reference": emotion_reference,
@@ -268,10 +289,10 @@ def vits_voice_conversion_api():
     if request.method == "POST":
         try:
             voice = request.files['upload']
-            original_id = int(request.form["original_id"])
-            target_id = int(request.form["target_id"])
-            format = request.form.get("format", voice.filename.split(".")[1])
-            use_streaming = request.form.get('streaming', False, type=bool)
+            original_id = get_param(request.form, "original_id", 0, int)
+            target_id = get_param(request.form, "target_id", 0, int)
+            format = get_param(request.form, "format", voice.filename.split(".")[1], str)
+            use_streaming = get_param(request.form, 'streaming', False, bool)
         except Exception as e:
             logger.error(f"[vits_voice_convertsion] {e}")
             return make_response("parameter error", 400)
@@ -345,7 +366,6 @@ def dimensional_emotion_api():
     if request.method == "POST":
         try:
             audio = request.files['upload']
-            use_streaming = request.form.get('streaming', False, type=bool)
         except Exception as e:
             logger.error(f"[dimensional_emotion] {e}")
             return make_response("parameter error", 400)
@@ -355,14 +375,7 @@ def dimensional_emotion_api():
     file_type = "application/octet-stream; charset=ascii"
     fname = os.path.splitext(audio.filename)[0] + ".npy"
     emotion_npy = tts_manager.get_dimensional_emotion_npy(content)
-    if use_streaming:
-        emotion_npy = tts_manager.generate_audio_chunks(emotion_npy)
-        response = make_response(emotion_npy)
-        response.headers['Content-Disposition'] = f'attachment; filename={fname}'
-        response.headers['Content-Type'] = file_type
-        return response
-    else:
-        return send_file(path_or_file=emotion_npy, mimetype=file_type, download_name=fname)
+    return send_file(path_or_file=emotion_npy, mimetype=file_type, download_name=fname)
 
 
 @voice_api.route('/bert-vits2', methods=["GET", "POST"])
@@ -378,16 +391,16 @@ def voice_bert_vits2_api():
             else:
                 request_data = request.form
 
-        text = request_data.get("text", "")
-        id = int(request_data.get("id", current_app.config.get("ID", 0)))
-        format = request_data.get("format", current_app.config.get("FORMAT", "wav"))
-        lang = request_data.get("lang", "auto").lower()
-        length = float(request_data.get("length", current_app.config.get("LENGTH", 1)))
-        noise = float(request_data.get("noise", current_app.config.get("NOISE", 0.667)))
-        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)
+        text = get_param(request_data, "text", "", str)
+        id = get_param(request_data, "id", current_app.config.get("ID", 0), int)
+        format = get_param(request_data, "format", current_app.config.get("FORMAT", "wav"), str)
+        lang = get_param(request_data, "lang", current_app.config.get("LANG", "auto"), str).lower()
+        length = get_param(request_data, "length", current_app.config.get("LENGTH", 1), float)
+        noise = get_param(request_data, "noise", current_app.config.get("NOISE", 0.667), float)
+        noisew = get_param(request_data, "noisew", current_app.config.get("NOISEW", 0.8), float)
+        sdp_ratio = get_param(request_data, "sdp_ratio", current_app.config.get("SDP_RATIO", 0.2), float)
+        segment_size = get_param(request_data, "segment_size", current_app.config.get("SEGMENT_SIZE", 50), int)
+        use_streaming = get_param(request_data, 'streaming', False, bool)
     except Exception as e:
         logger.error(f"[{ModelType.BERT_VITS2.value}] {e}")
         return make_response("parameter error", 400)
@@ -426,16 +439,16 @@ def voice_bert_vits2_api():
     fname = f"{str(uuid.uuid1())}.{format}"
     file_type = f"audio/{format}"
     state = {"text": text,
-            "id": id,
-            "format": format,
-            "length": length,
-            "noise": noise,
-            "noisew": noisew,
-            "sdp_ratio": sdp_ratio,
-            "max": max,
-            "lang": lang,
-            "speaker_lang": speaker_lang}
-    
+             "id": id,
+             "format": format,
+             "length": length,
+             "noise": noise,
+             "noisew": noisew,
+             "sdp_ratio": sdp_ratio,
+             "segment_size": segment_size,
+             "lang": lang,
+             "speaker_lang": speaker_lang}
+
     if use_streaming:
         audio = tts_manager.stream_bert_vits2_infer(state)
         response = make_response(audio)
@@ -448,8 +461,6 @@ def voice_bert_vits2_api():
         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}")
         path = os.path.join(current_app.config.get('CACHE_PATH'), fname)
@@ -505,10 +516,3 @@ def check():
     logger.info(f"[check] check id:{id} name:{name} lang:{lang}")
 
     return make_response(jsonify({"status": "success", "id": id, "name": name, "lang": lang}), 200)
-
-    # regular cleaning
-    @scheduler.task('interval', id='clean_task', seconds=current_app.config.get("CLEAN_INTERVAL_SECONDS", 3600),
-                    misfire_grace_time=900)
-    def clean_task():
-        clean_folder(current_app.config["UPLOAD_FOLDER"])
-        clean_folder(current_app.config["CACHE_PATH"])

utils/data_utils.py

@@ -82,3 +82,11 @@ def load_audio_to_torch(full_path, target_sampling_rate):
 def check_is_none(item) -> bool:
     # none -> True, not none -> False
     return item is None or (isinstance(item, str) and str(item).isspace()) or str(item) == ""
+
+
+def clean_folder(folder_path):
+    for filename in os.listdir(folder_path):
+        file_path = os.path.join(folder_path, filename)
+        # 如果是文件，则删除文件。如果是文件夹则跳过。
+        if os.path.isfile(file_path):
+            os.remove(file_path)