Spaces:
Configuration error
Configuration error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from gtts import gTTS
|
3 |
+
from moviepy.editor import TextClip, AudioFileClip
|
4 |
+
from transformers import RagTokenizer, RagRetriever, RagSequenceForGeneration
|
5 |
+
import torch
|
6 |
+
import tempfile
|
7 |
+
|
8 |
+
# Initialize RAG model components
|
9 |
+
tokenizer = RagTokenizer.from_pretrained("facebook/rag-sequence-nq")
|
10 |
+
retriever = RagRetriever.from_pretrained("facebook/rag-sequence-nq", index_name="exact", use_dummy_dataset=True)
|
11 |
+
model = RagSequenceForGeneration.from_pretrained("facebook/rag-sequence-nq", retriever=retriever)
|
12 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
13 |
+
model = model.to(device)
|
14 |
+
|
15 |
+
def generate_response(input_text):
|
16 |
+
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to(device)
|
17 |
+
generated = model.generate(input_ids)
|
18 |
+
response = tokenizer.batch_decode(generated, skip_special_tokens=True)[0]
|
19 |
+
return response
|
20 |
+
|
21 |
+
def text_to_speech(text):
|
22 |
+
tts = gTTS(text)
|
23 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix='.mp3') as temp_audio_file:
|
24 |
+
tts.save(temp_audio_file.name)
|
25 |
+
return temp_audio_file.name
|
26 |
+
|
27 |
+
def text_to_video(text, audio_filename):
|
28 |
+
text_clip = TextClip(text, fontsize=50, color='white', bg_color='black', size=(640, 480))
|
29 |
+
text_clip = text_clip.set_duration(10)
|
30 |
+
audio_clip = AudioFileClip(audio_filename)
|
31 |
+
video_clip = text_clip.set_audio(audio_clip)
|
32 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as temp_video_file:
|
33 |
+
video_clip.write_videofile(temp_video_file.name, codec='libx264')
|
34 |
+
return temp_video_file.name
|
35 |
+
|
36 |
+
def process_text(input_text):
|
37 |
+
response = generate_response(input_text)
|
38 |
+
audio_file = text_to_speech(response)
|
39 |
+
video_file = text_to_video(response, audio_file)
|
40 |
+
return response, audio_file, video_file
|
41 |
+
|
42 |
+
iface = gr.Interface(
|
43 |
+
fn=process_text,
|
44 |
+
inputs=gr.Textbox(label="Enter your text:"),
|
45 |
+
outputs=[gr.Textbox(label="RAG Model Response"), gr.Audio(label="Audio"), gr.Video(label="Video")],
|
46 |
+
live=True
|
47 |
+
)
|
48 |
+
|
49 |
+
iface.launch()
|