feat: use docker space

Dockerfile

@@ -0,0 +1,28 @@
+FROM python:3.11
+
+# By using XTTS you agree to CPML license https://coqui.ai/cpml
+ENV COQUI_TOS_AGREED=1
+
+# Set up a new user named "user" with user ID 1000
+RUN useradd -m -u 1000 user
+
+# Switch to the "user" user
+USER user
+
+# Set home to the user's home directory
+ENV HOME=/home/user \
+	PATH=/home/user/.local/bin:$PATH
+
+# Set the working directory to the user's home directory
+WORKDIR $HOME/app
+
+# Install dependencies
+COPY --chown=user:user requirements.txt .
+RUN pip install -r requirements.txt
+RUN python -m unidic download
+
+# Install model weights
+COPY --chown=user:user . .
+RUN python build.py
+
+CMD ["python", "app.py"]

README.md

@@ -3,10 +3,8 @@ title: XTTS
 emoji: 🐸
 colorFrom: green
 colorTo: red
-sdk: gradio
-sdk_version: 3.48.0
-app_file: app.py
 pinned: false
+sdk: docker
 models:
 - coqui/XTTS-v2
 ---

app.py

@@ -1,38 +1,26 @@
-import sys
-import io, os, stat
+import os
 import subprocess
-import random
-from zipfile import ZipFile
 import uuid
 import time
 import torch
 import torchaudio
 
-
-#download for mecab
-os.system('python -m unidic download')
-
-# By using XTTS you agree to CPML license https://coqui.ai/cpml
-os.environ["COQUI_TOS_AGREED"] = "1"
-
 # langid is used to detect language for longer text
 # Most users expect text to be their own language, there is checkbox to disable it
 import langid
-import base64
 import csv
 from io import StringIO
 import datetime
 import re
 
 import gradio as gr
-from scipy.io.wavfile import write
-from pydub import AudioSegment
 
-from TTS.api import TTS
 from TTS.tts.configs.xtts_config import XttsConfig
 from TTS.tts.models.xtts import Xtts
 from TTS.utils.generic_utils import get_user_data_dir
 
+print("application starting")
+
 HF_TOKEN = os.environ.get("HF_TOKEN")
 
 from huggingface_hub import HfApi
@@ -41,21 +29,10 @@ from huggingface_hub import HfApi
 api = HfApi(token=HF_TOKEN)
 repo_id = "coqui/xtts"
 
-# Use never ffmpeg binary for Ubuntu20 to use denoising for microphone input
-print("Export newer ffmpeg binary for denoise filter")
-ZipFile("ffmpeg.zip").extractall()
-print("Make ffmpeg binary executable")
-st = os.stat("ffmpeg")
-os.chmod("ffmpeg", st.st_mode | stat.S_IEXEC)
-
-# This will trigger downloading model
-print("Downloading if not downloaded Coqui XTTS V2")
-from TTS.utils.manage import ModelManager
+print("loading model")
 
 model_name = "tts_models/multilingual/multi-dataset/xtts_v2"
-ModelManager().download_model(model_name)
 model_path = os.path.join(get_user_data_dir("tts"), model_name.replace("/", "--"))
-print("XTTS downloaded")
 
 config = XttsConfig()
 config.load_json(os.path.join(model_path, "config.json"))
@@ -66,9 +43,15 @@ model.load_checkpoint(
     checkpoint_path=os.path.join(model_path, "model.pth"),
     vocab_path=os.path.join(model_path, "vocab.json"),
     eval=True,
-    use_deepspeed=True,
+    use_deepspeed=False,
 )
-model.cuda()
+
+if torch.cuda.is_available():
+    model.cuda()
+else:
+    model.cpu()
+
+print("Model loaded")
 
 # This is for debugging purposes only
 DEVICE_ASSERT_DETECTED = 0
@@ -81,8 +64,6 @@ def predict(
     prompt,
     language,
     audio_file_pth,
-    mic_file_path,
-    use_mic,
     voice_cleanup,
     no_lang_auto_detect,
     agree,
@@ -130,22 +111,7 @@ def predict(
                     None,
                 )
 
-        if use_mic == True:
-            if mic_file_path is not None:
-                speaker_wav = mic_file_path
-            else:
-                gr.Warning(
-                    "Please record your voice with Microphone, or uncheck Use Microphone to use reference audios"
-                )
-                return (
-                    None,
-                    None,
-                    None,
-                    None,
-                )
-
-        else:
-            speaker_wav = audio_file_pth
+        speaker_wav = audio_file_pth
 
         # Filtering for microphone input, as it has BG noise, maybe silence in beginning and end
         # This is fast filtering not perfect
@@ -328,8 +294,6 @@ def predict(
                     prompt,
                     language,
                     audio_file_pth,
-                    mic_file_path,
-                    use_mic,
                     voice_cleanup,
                     no_lang_auto_detect,
                     agree,
@@ -450,160 +414,6 @@ article = """
 <p>We collect data only for error cases for improvement.</p>
 </div>
 """
-examples = [
-    [
-        "Once when I was six years old I saw a magnificent picture",
-        "en",
-        "examples/female.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Lorsque j'avais six ans j'ai vu, une fois, une magnifique image",
-        "fr",
-        "examples/male.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Als ich sechs war, sah ich einmal ein wunderbares Bild",
-        "de",
-        "examples/female.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Cuando tenía seis años, vi una vez una imagen magnífica",
-        "es",
-        "examples/male.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Quando eu tinha seis anos eu vi, uma vez, uma imagem magnífica",
-        "pt",
-        "examples/female.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Kiedy miałem sześć lat, zobaczyłem pewnego razu wspaniały obrazek",
-        "pl",
-        "examples/male.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Un tempo lontano, quando avevo sei anni, vidi un magnifico disegno",
-        "it",
-        "examples/female.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Bir zamanlar, altı yaşındayken, muhteşem bir resim gördüm",
-        "tr",
-        "examples/female.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Когда мне было шесть лет, я увидел однажды удивительную картинку",
-        "ru",
-        "examples/female.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Toen ik een jaar of zes was, zag ik op een keer een prachtige plaat",
-        "nl",
-        "examples/male.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "Když mi bylo šest let, viděl jsem jednou nádherný obrázek",
-        "cs",
-        "examples/female.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "当我还只有六岁的时候， 看到了一副精彩的插画",
-        "zh-cn",
-        "examples/female.wav",
-        None,
-        False,
-        False,
-        False,
-        True,
-    ],
-    [
-        "かつて 六歳のとき、素晴らしい絵を見ました",
-        "ja",
-        "examples/female.wav",
-        None,
-        False,
-        True,
-        False,
-        True,
-    ],
-    [
-        "한번은 내가 여섯 살이었을 때 멋진 그림을 보았습니다.",
-        "ko",
-        "examples/female.wav",
-        None,
-        False,
-        True,
-        False,
-        True,
-    ],
-        [
-        "Egyszer hat éves koromban láttam egy csodálatos képet",
-        "hu",
-        "examples/male.wav",
-        None,
-        False,
-        True,
-        False,
-        True,
-    ],
-]
-
-
 
 with gr.Blocks(analytics_enabled=False) as demo:
     with gr.Row():
@@ -651,7 +461,6 @@ with gr.Blocks(analytics_enabled=False) as demo:
                     "ko",
                     "hu"
                 ],
-                max_choices=1,
                 value="en",
             )
             ref_gr = gr.Audio(
@@ -660,17 +469,6 @@ with gr.Blocks(analytics_enabled=False) as demo:
                 type="filepath",
                 value="examples/female.wav",
             )
-            mic_gr = gr.Audio(
-                source="microphone",
-                type="filepath",
-                info="Use your microphone to record audio",
-                label="Use Microphone for Reference",
-            )
-            use_mic_gr = gr.Checkbox(
-                label="Use Microphone",
-                value=False,
-                info="Notice: Microphone input may not work properly under traffic",
-            )
             clean_ref_gr = gr.Checkbox(
                 label="Cleanup Reference Voice",
                 value=False,
@@ -696,15 +494,7 @@ with gr.Blocks(analytics_enabled=False) as demo:
             out_text_gr = gr.Text(label="Metrics")
             ref_audio_gr = gr.Audio(label="Reference Audio Used")
 
-    with gr.Row():
-        gr.Examples(examples,
-                    label="Examples",
-                    inputs=[input_text_gr, language_gr, ref_gr, mic_gr, use_mic_gr, clean_ref_gr, auto_det_lang_gr, tos_gr],
-                    outputs=[video_gr, audio_gr, out_text_gr, ref_audio_gr],
-                    fn=predict,
-                    cache_examples=False,)
-
-    tts_button.click(predict, [input_text_gr, language_gr, ref_gr, mic_gr, use_mic_gr, clean_ref_gr, auto_det_lang_gr, tos_gr], outputs=[video_gr, audio_gr, out_text_gr, ref_audio_gr])
-
-demo.queue()  
-demo.launch(debug=True, show_api=True)
+    tts_button.click(predict, [input_text_gr, language_gr, ref_gr, clean_ref_gr, auto_det_lang_gr, tos_gr], outputs=[video_gr, audio_gr, out_text_gr, ref_audio_gr])
+ 
+print("Starting server")
+demo.queue().launch(debug=True, show_api=True)

build.py

@@ -0,0 +1,17 @@
+import os, stat
+from zipfile import ZipFile
+
+# Use never ffmpeg binary for Ubuntu20 to use denoising for microphone input
+print("Export newer ffmpeg binary for denoise filter")
+ZipFile("ffmpeg.zip").extractall()
+print("Make ffmpeg binary executable")
+st = os.stat("ffmpeg")
+os.chmod("ffmpeg", st.st_mode | stat.S_IEXEC)
+
+# This will trigger downloading model
+print("Downloading if not downloaded Coqui XTTS V2")
+from TTS.utils.manage import ModelManager
+
+model_name = "tts_models/multilingual/multi-dataset/xtts_v2"
+ModelManager().download_model(model_name)
+print("XTTS downloaded")

requirements.txt

@@ -8,5 +8,5 @@ mecab-python3==1.0.6
 unidic-lite==1.0.8
 unidic==1.1.0
 langid
-deepspeed
 pydub
+gradio