feat: file upload

GPT_SoVITS/inference_webui.py

@@ -39,6 +39,9 @@ import torch
 import torchaudio
 from text.LangSegmenter import LangSegmenter
 
+from scipy.io.wavfile import write
+import requests
+import io
 import zipfile
 from huggingface_hub import hf_hub_download
 import nltk
@@ -808,6 +811,7 @@ def get_tts_wav(
     sample_steps=8,
     if_sr=False,
     pause_second=0.3,
+    uploadParams=None
 ):
     global cache
     if ref_wav_path:
@@ -1052,7 +1056,21 @@ def get_tts_wav(
             audio_opt /= max_audio
     else:
         audio_opt = audio_opt.cpu().detach().numpy()
-    yield opt_sr, (audio_opt * 32767).astype(np.int16)
+    audioData = (audio_opt * 32767).astype(np.int16)
+    uploadAudio(opt_sr, audioData, uploadParams)
+    yield opt_sr, audioData
+
+
+def uploadAudio(opt_sr, audio_int16, uploadParams):
+    bio = io.BytesIO()
+    write(bio, opt_sr, audio_int16)
+    files = {
+        "file": ("audio.wav", bio.getvalue(), "audio/wav")
+    }
+    url = uploadParams['url']
+    del uploadParams['url']
+    response = requests.post(url, files=files, data=uploadParams)
+    print(response.text)
 
 
 def split(todo_text):
@@ -1216,8 +1234,11 @@ with gr.Blocks(title="GPT-SoVITS WebUI", analytics_enabled=False, js=js, css=css
         #  SoVITS_dropdown, GPT_dropdown])
         gr.Markdown(html_center(i18n("*请上传并填写参考信息"), "h3"))
         with gr.Row():
-            inp_ref = gr.Audio(label=i18n(
-                "请上传3~10秒内参考音频，超过会报错！"), type="filepath", scale=13)
+            with gr.Column():
+                inp_ref = gr.Audio(label=i18n(
+                    "请上传3~10秒内参考音频，超过会报错！"), type="filepath", scale=13)
+                uploadParams = gr.Textbox(label=i18n("结果上传参数"),
+                                          value="", lines=2, max_lines=2)
             with gr.Column(scale=13):
                 ref_text_free = gr.Checkbox(
                     label=i18n("开启无参考文本模式。不填参考文本亦相当于开启。")
@@ -1235,7 +1256,7 @@ with gr.Blocks(title="GPT-SoVITS WebUI", analytics_enabled=False, js=js, css=css
                     )
                 )
                 prompt_text = gr.Textbox(label=i18n(
-                    "参考音频的文本"), value="", lines=5, max_lines=5, scale=1)
+                    "参考音频的文本"), value="", lines=6, max_lines=6, scale=1)
             with gr.Column(scale=14):
                 prompt_language = gr.Dropdown(
                     label=i18n("参考音频的语种"),
@@ -1367,6 +1388,7 @@ with gr.Blocks(title="GPT-SoVITS WebUI", analytics_enabled=False, js=js, css=css
                 sample_steps,
                 if_sr_Checkbox,
                 pause_second_slider,
+                uploadParams
             ],
             [output],
         )