Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -4,69 +4,109 @@ from PIL import Image
|
|
| 4 |
import tempfile
|
| 5 |
from gradio_client import Client, handle_file
|
| 6 |
import torch
|
| 7 |
-
from transformers import VitsModel, AutoTokenizer
|
| 8 |
import scipy.io.wavfile as wavfile
|
| 9 |
import traceback
|
| 10 |
|
| 11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 13 |
print(f"Using device: {device}")
|
| 14 |
|
| 15 |
try:
|
| 16 |
-
# TTS модель
|
| 17 |
tts_model = VitsModel.from_pretrained("facebook/mms-tts-kaz").to(device)
|
| 18 |
tts_tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-kaz")
|
| 19 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
except Exception as e:
|
| 21 |
-
raise RuntimeError(f"Ошибка загрузки
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
-
# Space для talking-head
|
| 24 |
TALKING_HEAD_SPACE = "Skywork/skyreels-a1-talking-head"
|
| 25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
def inference(image: Image.Image, text: str):
|
|
|
|
| 27 |
error_msg = ""
|
| 28 |
video_path = None
|
| 29 |
audio_path = None
|
| 30 |
img_path = None
|
|
|
|
| 31 |
try:
|
| 32 |
-
#
|
| 33 |
if image is None:
|
| 34 |
raise ValueError("Загрузите изображение лектора!")
|
|
|
|
| 35 |
if not text or not text.strip():
|
| 36 |
raise ValueError("Введите текст лекции!")
|
|
|
|
| 37 |
if len(text) > 500:
|
| 38 |
-
raise ValueError("Текст
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
with torch.no_grad():
|
| 44 |
output = tts_model(**inputs)
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
raise ValueError("TTS сгенерировал пустое аудио! Попробуйте другой текст.")
|
| 48 |
-
# Конвертация в int16 для WAV
|
| 49 |
audio = (waveform * 32767).astype("int16")
|
| 50 |
sampling_rate = tts_model.config.sampling_rate
|
| 51 |
-
|
| 52 |
-
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as
|
| 53 |
-
wavfile.write(
|
| 54 |
-
audio_path =
|
| 55 |
-
|
| 56 |
-
#
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
|
|
|
|
|
|
| 66 |
client = Client(TALKING_HEAD_SPACE)
|
| 67 |
-
|
| 68 |
-
print("Доступные API методы:", client.view_api())
|
| 69 |
-
# Вызов API с правильными параметрами
|
| 70 |
result = client.predict(
|
| 71 |
image_path=handle_file(img_path),
|
| 72 |
audio_path=handle_file(audio_path),
|
|
@@ -74,67 +114,63 @@ def inference(image: Image.Image, text: str):
|
|
| 74 |
steps=10,
|
| 75 |
api_name="/process_image_audio"
|
| 76 |
)
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
video_data = result[0]
|
| 81 |
-
if isinstance(video_data, dict) and 'video' in video_data:
|
| 82 |
-
video_path = video_data['video']
|
| 83 |
-
elif isinstance(video_data, dict) and 'path' in video_data:
|
| 84 |
-
video_path = video_data['path']
|
| 85 |
-
elif isinstance(video_data, str):
|
| 86 |
-
video_path = video_data
|
| 87 |
-
else:
|
| 88 |
-
video_path = video_data
|
| 89 |
else:
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
error_msg = "✅ Видео успешно
|
|
|
|
| 93 |
except Exception as e:
|
| 94 |
error_msg = f"❌ Ошибка: {str(e)}"
|
| 95 |
-
print(f"ОШИБКА: {error_msg}")
|
| 96 |
traceback.print_exc()
|
|
|
|
| 97 |
finally:
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
if img_path and os.path.exists(img_path):
|
| 106 |
-
try:
|
| 107 |
-
os.remove(img_path)
|
| 108 |
-
print(f"Удален временный файл: {img_path}")
|
| 109 |
-
except:
|
| 110 |
-
pass
|
| 111 |
return video_path, error_msg
|
| 112 |
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 119 |
|
| 120 |
iface = gr.Interface(
|
| 121 |
fn=inference,
|
| 122 |
inputs=[
|
| 123 |
-
gr.Image(type="pil", label="📸 Фото
|
| 124 |
gr.Textbox(
|
| 125 |
lines=5,
|
| 126 |
-
|
| 127 |
-
|
| 128 |
)
|
| 129 |
],
|
| 130 |
outputs=[
|
| 131 |
-
gr.Video(label="🎬
|
| 132 |
-
gr.Textbox(label="ℹ️
|
| 133 |
],
|
| 134 |
title=title,
|
| 135 |
description=description,
|
| 136 |
-
cache_examples=False
|
|
|
|
| 137 |
)
|
| 138 |
|
| 139 |
if __name__ == "__main__":
|
| 140 |
-
iface.launch()
|
|
|
|
| 4 |
import tempfile
|
| 5 |
from gradio_client import Client, handle_file
|
| 6 |
import torch
|
| 7 |
+
from transformers import VitsModel, AutoTokenizer, pipeline
|
| 8 |
import scipy.io.wavfile as wavfile
|
| 9 |
import traceback
|
| 10 |
|
| 11 |
+
|
| 12 |
+
# =========================
|
| 13 |
+
# Загрузка моделей
|
| 14 |
+
# =========================
|
| 15 |
+
|
| 16 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 17 |
print(f"Using device: {device}")
|
| 18 |
|
| 19 |
try:
|
| 20 |
+
# TTS модель казахского языка
|
| 21 |
tts_model = VitsModel.from_pretrained("facebook/mms-tts-kaz").to(device)
|
| 22 |
tts_tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-kaz")
|
| 23 |
+
|
| 24 |
+
# Модель перевода ru -> kk
|
| 25 |
+
translator = pipeline(
|
| 26 |
+
"translation",
|
| 27 |
+
model="facebook/nllb-200-distilled-600M",
|
| 28 |
+
device=0 if device == "cuda" else -1
|
| 29 |
+
)
|
| 30 |
+
|
| 31 |
+
print("✅ Все модели успешно загружены!")
|
| 32 |
+
|
| 33 |
except Exception as e:
|
| 34 |
+
raise RuntimeError(f"Ошибка загрузки моделей: {str(e)}")
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
# =========================
|
| 38 |
+
# Talking Head Space
|
| 39 |
+
# =========================
|
| 40 |
|
|
|
|
| 41 |
TALKING_HEAD_SPACE = "Skywork/skyreels-a1-talking-head"
|
| 42 |
|
| 43 |
+
|
| 44 |
+
# =========================
|
| 45 |
+
# Основная функция
|
| 46 |
+
# =========================
|
| 47 |
+
|
| 48 |
def inference(image: Image.Image, text: str):
|
| 49 |
+
|
| 50 |
error_msg = ""
|
| 51 |
video_path = None
|
| 52 |
audio_path = None
|
| 53 |
img_path = None
|
| 54 |
+
|
| 55 |
try:
|
| 56 |
+
# Проверки
|
| 57 |
if image is None:
|
| 58 |
raise ValueError("Загрузите изображение лектора!")
|
| 59 |
+
|
| 60 |
if not text or not text.strip():
|
| 61 |
raise ValueError("Введите текст лекции!")
|
| 62 |
+
|
| 63 |
if len(text) > 500:
|
| 64 |
+
raise ValueError("Текст превышает 500 символов!")
|
| 65 |
+
|
| 66 |
+
print("Ввод (RU):", text)
|
| 67 |
+
|
| 68 |
+
# =========================
|
| 69 |
+
# Шаг 1 — Перевод
|
| 70 |
+
# =========================
|
| 71 |
+
translation = translator(
|
| 72 |
+
text,
|
| 73 |
+
src_lang="rus_Cyrl",
|
| 74 |
+
tgt_lang="kaz_Cyrl"
|
| 75 |
+
)
|
| 76 |
+
|
| 77 |
+
translated_text = translation[0]["translation_text"]
|
| 78 |
+
print("Перевод (KK):", translated_text)
|
| 79 |
+
|
| 80 |
+
# =========================
|
| 81 |
+
# Шаг 2 — Озвучка
|
| 82 |
+
# =========================
|
| 83 |
+
inputs = tts_tokenizer(translated_text, return_tensors="pt").to(device)
|
| 84 |
+
|
| 85 |
with torch.no_grad():
|
| 86 |
output = tts_model(**inputs)
|
| 87 |
+
|
| 88 |
+
waveform = output.waveform.squeeze().cpu().numpy()
|
|
|
|
|
|
|
| 89 |
audio = (waveform * 32767).astype("int16")
|
| 90 |
sampling_rate = tts_model.config.sampling_rate
|
| 91 |
+
|
| 92 |
+
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
|
| 93 |
+
wavfile.write(f.name, sampling_rate, audio)
|
| 94 |
+
audio_path = f.name
|
| 95 |
+
|
| 96 |
+
# =========================
|
| 97 |
+
# Шаг 3 — Сохранение изображения
|
| 98 |
+
# =========================
|
| 99 |
+
with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as f:
|
| 100 |
+
if image.mode != "RGB":
|
| 101 |
+
image = image.convert("RGB")
|
| 102 |
+
image.save(f.name)
|
| 103 |
+
img_path = f.name
|
| 104 |
+
|
| 105 |
+
# =========================
|
| 106 |
+
# Шаг 4 — Генерация видео
|
| 107 |
+
# =========================
|
| 108 |
client = Client(TALKING_HEAD_SPACE)
|
| 109 |
+
|
|
|
|
|
|
|
| 110 |
result = client.predict(
|
| 111 |
image_path=handle_file(img_path),
|
| 112 |
audio_path=handle_file(audio_path),
|
|
|
|
| 114 |
steps=10,
|
| 115 |
api_name="/process_image_audio"
|
| 116 |
)
|
| 117 |
+
|
| 118 |
+
if isinstance(result, tuple):
|
| 119 |
+
video_path = result[0]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 120 |
else:
|
| 121 |
+
raise ValueError("Видео не получено!")
|
| 122 |
+
|
| 123 |
+
error_msg = "✅ Видео успешно создано!"
|
| 124 |
+
|
| 125 |
except Exception as e:
|
| 126 |
error_msg = f"❌ Ошибка: {str(e)}"
|
|
|
|
| 127 |
traceback.print_exc()
|
| 128 |
+
|
| 129 |
finally:
|
| 130 |
+
for p in [audio_path, img_path]:
|
| 131 |
+
if p and os.path.exists(p):
|
| 132 |
+
try:
|
| 133 |
+
os.remove(p)
|
| 134 |
+
except:
|
| 135 |
+
pass
|
| 136 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 137 |
return video_path, error_msg
|
| 138 |
|
| 139 |
+
|
| 140 |
+
# =========================
|
| 141 |
+
# Gradio Интерфейс
|
| 142 |
+
# =========================
|
| 143 |
+
|
| 144 |
+
title = "Бейне Оқытушы"
|
| 145 |
+
|
| 146 |
+
description = """
|
| 147 |
+
Суретіңізді жүктеп, дәріс мәтінін орыс тілінде енгізіңіз.
|
| 148 |
+
Жүйе автоматты түрде қазақ тіліне аударады және бейне жасайды!
|
| 149 |
+
|
| 150 |
+
**Талаптар:**
|
| 151 |
+
- Фото: бет анық көрінетін
|
| 152 |
+
- Мәтін: орыс тілінде (500 таңбаға дейін)
|
| 153 |
+
"""
|
| 154 |
|
| 155 |
iface = gr.Interface(
|
| 156 |
fn=inference,
|
| 157 |
inputs=[
|
| 158 |
+
gr.Image(type="pil", label="📸 Фото дәріскер"),
|
| 159 |
gr.Textbox(
|
| 160 |
lines=5,
|
| 161 |
+
label="📝 Дәріс мәтіні (орыс тілінде)",
|
| 162 |
+
placeholder="500 таңбаға дейін..."
|
| 163 |
)
|
| 164 |
],
|
| 165 |
outputs=[
|
| 166 |
+
gr.Video(label="🎬 Дайын бейне"),
|
| 167 |
+
gr.Textbox(label="ℹ️ Мәртебе")
|
| 168 |
],
|
| 169 |
title=title,
|
| 170 |
description=description,
|
| 171 |
+
cache_examples=False,
|
| 172 |
+
flagging_mode="never"
|
| 173 |
)
|
| 174 |
|
| 175 |
if __name__ == "__main__":
|
| 176 |
+
iface.launch()
|