Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Freddy Boulton
commited on
Commit
·
740c813
1
Parent(s):
a8b2330
add code
Browse files- app.py +79 -0
- requirements.txt +4 -0
app.py
ADDED
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import assemblyai as aai
|
2 |
+
import gradio as gr
|
3 |
+
import os
|
4 |
+
import subprocess
|
5 |
+
import uuid
|
6 |
+
from google import genai
|
7 |
+
from dotenv import load_dotenv
|
8 |
+
|
9 |
+
load_dotenv()
|
10 |
+
|
11 |
+
client = genai.Client(api_key=os.getenv("GEMINI_API_KEY"))
|
12 |
+
|
13 |
+
|
14 |
+
aai.settings.api_key = os.getenv("ASSEMBLYAI_API_KEY")
|
15 |
+
|
16 |
+
|
17 |
+
|
18 |
+
def _extract_audio_to_mp3(video_file: str) -> str:
|
19 |
+
"""Extract audio from an mp4 video file to a uniquely named mp3 file using ffmpeg."""
|
20 |
+
unique_id = uuid.uuid4().hex
|
21 |
+
mp3_file = f"audio_{unique_id}.mp3"
|
22 |
+
command = [
|
23 |
+
"ffmpeg",
|
24 |
+
"-i", video_file,
|
25 |
+
"-vn",
|
26 |
+
"-acodec", "libmp3lame",
|
27 |
+
"-y",
|
28 |
+
mp3_file
|
29 |
+
]
|
30 |
+
subprocess.run(command, check=True)
|
31 |
+
return mp3_file
|
32 |
+
|
33 |
+
|
34 |
+
def get_transcript(video_file: str) -> str:
|
35 |
+
"""Get a transcript of a video file using assembly ai API.
|
36 |
+
Arguments:
|
37 |
+
video_file: str - The path to the video file to transcribe.
|
38 |
+
Returns:
|
39 |
+
transcript: str - The transcript of the video file.
|
40 |
+
"""
|
41 |
+
try:
|
42 |
+
audio_file = _extract_audio_to_mp3(video_file)
|
43 |
+
config = aai.TranscriptionConfig(speech_model=aai.SpeechModel.best)
|
44 |
+
transcript = aai.Transcriber(config=config).transcribe(audio_file)
|
45 |
+
return transcript.text or "Could not transcribe video."
|
46 |
+
except Exception as e:
|
47 |
+
raise gr.Error(f"Error transcribing video: {e}")
|
48 |
+
|
49 |
+
transcript_iface = gr.Interface(fn=get_transcript, inputs=gr.Video(),
|
50 |
+
outputs="text")
|
51 |
+
|
52 |
+
|
53 |
+
def get_initial_post(transcript: str) -> str:
|
54 |
+
|
55 |
+
response = client.models.generate_content(
|
56 |
+
model="gemini-2.5-flash",
|
57 |
+
contents=f"""
|
58 |
+
You are an expert social media copywriter.
|
59 |
+
You are given a transcript of a video and you need to write a post for social media.
|
60 |
+
The post should be 280 characters or less so that it can be posted on Twitter without being truncated in the preview.
|
61 |
+
The post should include a catchy title that will grab the viewer's attention.
|
62 |
+
After the title, include some bullet points that will help the viewer understand the main points of the video.
|
63 |
+
Here is the transcript of a video:
|
64 |
+
{transcript}
|
65 |
+
""",
|
66 |
+
)
|
67 |
+
|
68 |
+
return response.text or "Could not generate post."
|
69 |
+
|
70 |
+
|
71 |
+
post_iface = gr.Interface(fn=get_initial_post, inputs=gr.Textbox(),
|
72 |
+
outputs=gr.Textbox())
|
73 |
+
|
74 |
+
|
75 |
+
iface = gr.TabbedInterface([transcript_iface, post_iface], ["Transcript", "Post"])
|
76 |
+
|
77 |
+
|
78 |
+
if __name__ == "__main__":
|
79 |
+
iface.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
assemblyai
|
2 |
+
google-genai
|
3 |
+
gradio
|
4 |
+
python-dotenv
|