updates

app.py

@@ -117,6 +117,7 @@ def tts(
     f0_method,
     index_rate,
     protect,
+    custom_audio,
     filter_radius=3,
     resample_sr=0,
     rms_mix_rate=0.25,
@@ -141,11 +142,16 @@ def tts(
             speed_str = f"+{speed}%"
         else:
             speed_str = f"{speed}%"
-        asyncio.run(
-            edge_tts.Communicate(
-                tts_text, "-".join(tts_voice.split("-")[:-1]), rate=speed_str
-            ).save(edge_output_filename)
-        )
+
+        if custom_audio is not None:
+            hertz, audio = custom_audio
+            edge_output_filename = audio
+        else:
+            asyncio.run(
+                edge_tts.Communicate(
+                    tts_text, "-".join(tts_voice.split("-")[:-1]), rate=speed_str
+                ).save(edge_output_filename)
+            )
         t1 = time.time()
         edge_time = t1 - t0
         audio, sr = librosa.load(edge_output_filename, sr=16000, mono=True)
@@ -217,16 +223,7 @@ rmvpe_model = RMVPE("rmvpe.pt", config.is_half, config.device)
 print("rmvpe model loaded.")
 
 initial_md = """
-# RVC text-to-speech demo
-
-This is a text-to-speech demo of RVC moe models of [rvc_okiba](https://huggingface.co/litagin/rvc_okiba) using [edge-tts](https://github.com/rany2/edge-tts).
-
-Input text ➡[(edge-tts)](https://github.com/rany2/edge-tts)➡ Speech mp3 file ➡[(RVC)](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)➡ Final output
-
-This runs on the 🤗 server's cpu, so it may be slow.
-
-Although the models are trained on Japanese voices and intended for Japanese text, they can also be used with other languages with the corresponding edge-tts speaker (but possibly with a Japanese accent).
-
+#
 Input characters are limited to 280 characters, and the speech audio is limited to 20 seconds in this 🤗 space.
 
 [Visit this GitHub repo](https://github.com/litagin02/rvc-tts-webui) for running locally with your models and GPU!
@@ -238,7 +235,7 @@ with app:
     with gr.Row():
         with gr.Column():
             model_name = gr.Dropdown(
-                label="Model (all models except man-_ are girl models)",
+                label="Model",
                 choices=models,
                 value=models[0],
             )
@@ -284,8 +281,9 @@ with app:
                 step=10,
                 interactive=True,
             )
-            tts_text = gr.Textbox(label="Input Text", value="これは日本語テキストから音声への変換デモです。")
+            tts_text = gr.Textbox(label="Input Text", value="Input text here")
         with gr.Column():
+            custom_audio = gr.Audio(label="Custom audio", type="filepath")
             but0 = gr.Button("Convert", variant="primary")
             info_text = gr.Textbox(label="Output info")
         with gr.Column():
@@ -302,6 +300,7 @@ with app:
                 f0_method,
                 index_rate,
                 protect0,
+                custom_audio,
             ],
             [info_text, edge_tts_output, tts_output],
         )

weights/Cyprien400/Cyprien_e400_s37600.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b5770b07e7dc0784dbd488c98f1a3e46ffec10de7773d183afda97444dc9257
+size 55228787

weights/Cyprien400/added_IVF1510_Flat_nprobe_1_Cyprien_v2.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:116f12bbb167f6d8029cb6ffe96d86a2bd75abd43f86fe714838ad345bde21c1
+size 186041379

weights/Inoxtag/added_IVF829_Flat_nprobe_1_inoxtagModel_v2.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee00965eb64990a80a8befd25efd7c65b9bcbab3d7e266c60b15b3e061491065
+size 102222259

weights/Inoxtag/inoxtagModel.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffb79ffebd382df21b7b0a06df491ea9ec9dd64bc43d0e0356cc3f4e20c4d5d6
+size 55225574

weights/Joyca/JOYCA (1).pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1373a35ed990141e2005d4c2b3d97ea7f48b60e44c26d7626bfd46cee81d44cc
+size 55193241

weights/Joyca/added_IVF854_Flat_nprobe_1_JOYCA_v2.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86e79e43e4946e94b8f662357b5f13c66bc4a7debc81c3285e016a496f85641c
+size 105268379

weights/LinedMC/metadata.json

@@ -0,0 +1,148 @@
+{
+  "title": "Lined (YTB Minecraft)",
+  "author": {
+    "name": "smaillor",
+    "discordUserId": "835772122303299596"
+  },
+  "md5": "173d23712a3d83d954d3a3c79d1f4071",
+  "uploadedAt": "2023-09-17T09:26:41.179Z",
+  "weightsLink": "https://www.weights.gg/models/clmn97p0r000hwszfs86k11aw",
+  "id": "clmn97p0r000hwszfs86k11aw",
+  "type": "v2",
+  "tags": [
+    "RVC v2",
+    "Jeux Vidéo",
+    "YouTubeur",
+    "Français"
+  ],
+  "description": "__Fait par :__ Smaillor\n__Lien HuggingFace :__ https://huggingface.co/Smaillor/Lined/resolve/main/Lined.zip",
+  "samples": [],
+  "files": [
+    {
+      "name": "model.index",
+      "size": 51269819,
+      "md5": "cb13f634f5f692a28e3e81c6625b2916"
+    },
+    {
+      "name": "model.pth",
+      "size": 57552537,
+      "md5": "173d23712a3d83d954d3a3c79d1f4071"
+    }
+  ],
+  "torchMetadata": {
+    "config": {
+      "spec_channels": 1025,
+      "segment_size": 32,
+      "inter_channels": 192,
+      "hidden_channels": 192,
+      "filter_channels": 768,
+      "n_heads": 2,
+      "n_layers": 6,
+      "kernel_size": 3,
+      "p_dropout": 0,
+      "resblock": "1",
+      "resblock_kernel_sizes": [
+        3,
+        7,
+        11
+      ],
+      "resblock_dilation_sizes": [
+        [
+          1,
+          3,
+          5
+        ],
+        [
+          1,
+          3,
+          5
+        ],
+        [
+          1,
+          3,
+          5
+        ]
+      ],
+      "upsample_rates": [
+        12,
+        10,
+        2,
+        2
+      ],
+      "upsample_initial_channel": 512,
+      "upsample_kernel_sizes": [
+        24,
+        20,
+        4,
+        4
+      ],
+      "emb_channels": null,
+      "spk_embed_dim": 109,
+      "gin_channels": 256,
+      "sr": 48000
+    },
+    "f0": 1,
+    "version": "v2",
+    "extra_info": {
+      "config": [
+        1025,
+        32,
+        192,
+        192,
+        768,
+        2,
+        6,
+        3,
+        0,
+        "1",
+        [
+          3,
+          7,
+          11
+        ],
+        [
+          [
+            1,
+            3,
+            5
+          ],
+          [
+            1,
+            3,
+            5
+          ],
+          [
+            1,
+            3,
+            5
+          ]
+        ],
+        [
+          12,
+          10,
+          2,
+          2
+        ],
+        512,
+        [
+          24,
+          20,
+          4,
+          4
+        ],
+        109,
+        256,
+        48000
+      ],
+      "info": "300epoch",
+      "sr": "48k",
+      "f0": 1,
+      "version": "v2"
+    }
+  },
+  "url": "https://huggingface.co/Smaillor/Lined/resolve/main/Lined.zip",
+  "originalFileList": [
+    "Lined.pth",
+    "added_IVF416_Flat_nprobe_1_Lined_v2.index"
+  ]
+}

weights/LinedMC/model.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73dd191e673ebaaba8a5b8f4a501e61a041d6889b0b443bff0213e3df8b03246
+size 51269819

weights/LinedMC/model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3da545f0b6d6e6a7f07d769a20830c1f6c8099e84a8de1893992ecd8c421d38
+size 57552537

weights/Michou/Michoumodel.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b0a15788014b5679f2fddce05a0fa08941b34be6c355e03a0c1c629da68e227
+size 55225115

weights/Michou/added_IVF1218_Flat_nprobe_1_Michoumodel_v2.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f978e9086f56ee411459fc90fc425f04c67277cbfcfe651405fe9fd7c3f090f
+size 150066979

weights/MonsieurPof/MonsieurPof.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffe29750969a4ce472f48176186030429c9146ce36f8457067004a1699e03add
+size 57589524

weights/MonsieurPof/added_IVF811_Flat_nprobe_1_MonsieurPof_v2.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53e559ebc48de8235313eb4c6fdb93d693134f9b8639d6c4dfac185b5fb24be7
+size 99986179

weights/Potatoz/added_IVF1049_Flat_nprobe_1_potatoz_v2.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:404909c3513da58418bc3f66670a82c69769020a2bdd953853d1fef255926b32
+size 129347819

weights/Potatoz/potatoz_e300_s29400.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb789ca44646f335a23cc12350cac348f74feaadcdced006381f572afc6f60b4
+size 55228787

weights/Psyhodelik/Psyhodelik.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bcb049e18a2a27cde3a380df33c21ac7c9a94cfbd93237c6f34ce9ec7e4cad3
+size 57583952

weights/Psyhodelik/added_IVF1795_Flat_nprobe_1_Psyhodelik_v2.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ec1e2af6278b521ce323efd2ff19cbe2a7d7b031d1e7dba4657d6856cd8dbaa
+size 221165699

weights/Shawen/metadata.json

@@ -0,0 +1,151 @@
+{
+  "title": "Shawen (YTB Minecraft) - RVC 2 - 300 EPOCHS",
+  "author": {
+    "name": "smaillor",
+    "discordUserId": "835772122303299596"
+  },
+  "md5": "3449250a9e23d3ce3b313c1cf9c82665",
+  "uploadedAt": "2023-11-06T22:08:21.681Z",
+  "weightsLink": "https://www.weights.gg/models/clonggc1u00rowswvikx53om1",
+  "id": "clonggc1u00rowswvikx53om1",
+  "type": "v2",
+  "tags": [
+    "RVC v2",
+    "Jeux Vidéo",
+    "YouTubeur",
+    "Français"
+  ],
+  "description": "__Fait par :__ Smaillor\n__Lien HuggingFace :__ https://huggingface.co/Smaillor/Shawen/resolve/main/Shawen.zip",
+  "samples": [],
+  "files": [
+    {
+      "name": "model.index",
+      "size": 50259579,
+      "md5": "caf60adfc7e7193f3b55e28f05057346"
+    },
+    {
+      "name": "model.pth",
+      "size": 57583034,
+      "md5": "3449250a9e23d3ce3b313c1cf9c82665"
+    }
+  ],
+  "torchMetadata": {
+    "config": {
+      "spec_channels": 1025,
+      "segment_size": 32,
+      "inter_channels": 192,
+      "hidden_channels": 192,
+      "filter_channels": 768,
+      "n_heads": 2,
+      "n_layers": 6,
+      "kernel_size": 3,
+      "p_dropout": 0,
+      "resblock": "1",
+      "resblock_kernel_sizes": [
+        3,
+        7,
+        11
+      ],
+      "resblock_dilation_sizes": [
+        [
+          1,
+          3,
+          5
+        ],
+        [
+          1,
+          3,
+          5
+        ],
+        [
+          1,
+          3,
+          5
+        ]
+      ],
+      "upsample_rates": [
+        12,
+        10,
+        2,
+        2
+      ],
+      "upsample_initial_channel": 512,
+      "upsample_kernel_sizes": [
+        24,
+        20,
+        4,
+        4
+      ],
+      "emb_channels": null,
+      "spk_embed_dim": 109,
+      "gin_channels": 256,
+      "sr": 48000
+    },
+    "f0": 1,
+    "version": "v2",
+    "extra_info": {
+      "config": [
+        1025,
+        32,
+        192,
+        192,
+        768,
+        2,
+        6,
+        3,
+        0,
+        "1",
+        [
+          3,
+          7,
+          11
+        ],
+        [
+          [
+            1,
+            3,
+            5
+          ],
+          [
+            1,
+            3,
+            5
+          ],
+          [
+            1,
+            3,
+            5
+          ]
+        ],
+        [
+          12,
+          10,
+          2,
+          2
+        ],
+        512,
+        [
+          24,
+          20,
+          4,
+          4
+        ],
+        109,
+        256,
+        48000
+      ],
+      "info": "300epoch",
+      "sr": "48k",
+      "f0": 1,
+      "version": "v2"
+    }
+  },
+  "url": "https://huggingface.co/Smaillor/Shawen/resolve/main/Shawen.zip",
+  "urls": [
+    "https://huggingface.co/Smaillor/Shawen/resolve/main/Shawen.zip"
+  ],
+  "originalFileList": [
+    "ShawenBG.pth",
+    "added_IVF407_Flat_nprobe_1_ShawenBG_v2.index"
+  ]
+}

weights/Shawen/model.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49da081047dcd0a0eb3b98d66d55a688e7ae5ba4b7f23e0e878eb9577e283a7e
+size 50259579

weights/Shawen/model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38b35db83233dd8df7c8fcb0b4c8e6383e4ba838221ad7d5680552fac472ccad
+size 57583034

weights/XJP/added_IVF336_Flat_nprobe_1.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd147ef8fe1a29616cd0ef342a80602bacd073e2a78e9a34243210c5a87a3ebc
+size 13886731

weights/XJP/xi-jingpin2333333.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd6d56ca24a360d93655f339fabf38180dd5d23625166ab26f4640ceb70af8a9
+size 55028048

weights/azpaz/metadata.json

@@ -0,0 +1,148 @@
+{
+  "title": "AzpazTv (YTB Minecraft)",
+  "author": {
+    "name": "smaillor",
+    "discordUserId": "835772122303299596"
+  },
+  "md5": "5c2b558e26d3ddcb41355370ec60fd23",
+  "uploadedAt": "2023-09-13T14:59:06.628Z",
+  "weightsLink": "https://www.weights.gg/models/clmjwhlgr0005wsunjqx5gfj7",
+  "id": "clmjwhlgr0005wsunjqx5gfj7",
+  "type": "v2",
+  "tags": [
+    "RVC v1",
+    "Jeux Vidéo",
+    "YouTubeur",
+    "Français"
+  ],
+  "description": "__Fait par :__ Smaillor\n__Conseil :__ Rglez le Pitch selon votre voix, sinon a sera mal fait\n__Lien HuggingFace :__ https://huggingface.co/Smaillor/AzpazTv/resolve/main/AzpazTv.zip",
+  "samples": [],
+  "files": [
+    {
+      "name": "model.index",
+      "size": 51263659,
+      "md5": "c01177580f931054afcb3fb5c14c22ea"
+    },
+    {
+      "name": "model.pth",
+      "size": 57581999,
+      "md5": "5c2b558e26d3ddcb41355370ec60fd23"
+    }
+  ],
+  "torchMetadata": {
+    "config": {
+      "spec_channels": 1025,
+      "segment_size": 32,
+      "inter_channels": 192,
+      "hidden_channels": 192,
+      "filter_channels": 768,
+      "n_heads": 2,
+      "n_layers": 6,
+      "kernel_size": 3,
+      "p_dropout": 0,
+      "resblock": "1",
+      "resblock_kernel_sizes": [
+        3,
+        7,
+        11
+      ],
+      "resblock_dilation_sizes": [
+        [
+          1,
+          3,
+          5
+        ],
+        [
+          1,
+          3,
+          5
+        ],
+        [
+          1,
+          3,
+          5
+        ]
+      ],
+      "upsample_rates": [
+        12,
+        10,
+        2,
+        2
+      ],
+      "upsample_initial_channel": 512,
+      "upsample_kernel_sizes": [
+        24,
+        20,
+        4,
+        4
+      ],
+      "emb_channels": null,
+      "spk_embed_dim": 109,
+      "gin_channels": 256,
+      "sr": 48000
+    },
+    "f0": 1,
+    "version": "v2",
+    "extra_info": {
+      "config": [
+        1025,
+        32,
+        192,
+        192,
+        768,
+        2,
+        6,
+        3,
+        0,
+        "1",
+        [
+          3,
+          7,
+          11
+        ],
+        [
+          [
+            1,
+            3,
+            5
+          ],
+          [
+            1,
+            3,
+            5
+          ],
+          [
+            1,
+            3,
+            5
+          ]
+        ],
+        [
+          12,
+          10,
+          2,
+          2
+        ],
+        512,
+        [
+          24,
+          20,
+          4,
+          4
+        ],
+        109,
+        256,
+        48000
+      ],
+      "info": "300epoch",
+      "sr": "48k",
+      "f0": 1,
+      "version": "v2"
+    }
+  },
+  "url": "https://huggingface.co/Smaillor/AzpazTv/resolve/main/AzpazTv.zip",
+  "originalFileList": [
+    "AzpazTv.pth",
+    "added_IVF416_Flat_nprobe_1_AzpazTv_v2.index"
+  ]
+}

weights/azpaz/model.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aade5967215beed1a8fbb43b0db5d3104aeeccf80d8cf19449c6da81f1714d63
+size 51263659

weights/azpaz/model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b83e4cc305f3e6acace8dc012ca08f5c49878c0bc204a412bc028acce8cfffc
+size 57581999

weights/lincn/lincn_e420_s7560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cbeb4bce326fe9042f80c51b0cea29eb0526dabf03cbbbff2fd3e2179403690
+size 55227861

weights/lincn/trained_IVF718_Flat_nprobe_1_lincn_v2.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5143863f8bbe8e0ebc4a07aee282d3b3048e16c2cc0dec8d947f622d97d196fc
+size 2205835