subtitile / README.md
mrpoddaa's picture
Upload 3 files
1c800d7 verified
|
Raw
History Blame Contribute Delete
1.66 kB
---
title: Sinhala SRT Subtitle Cleaner
emoji: 🎬
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: "4.44.0"
app_file: app.py
pinned: false
license: mit
---
# 🎬 සිංහල SRT Subtitle Cleaner
Gemma 2 9B AI model භාවිතා කර .srt subtitle files වලින් author credits, website URLs, සහ copyright notices ස්වයංක්‍රීයව ඉවත් කරයි.
## ✨ Features
- **Dual Detection:** Regex fast-filter + Gemma 2 9B AI classification
- **Live Interface:** File upload හෝ text paste කළ හැකිය
- **API Endpoint:** Python/cURL හරහා programmatic access
- **Renumbering:** ඉවත් කළ lines වලින් පසු automatic renumbering
- **Sinhala Support:** UTF-8 encoding සම්පූර්ණයෙන් support කරයි
## 🚀 Setup
### HF Token
Space Settings > Variables and Secrets හි:
```
HF_TOKEN = hf_xxxxxxxxxxxxxxxx
```
Gemma 2 access සඳහා [model page](https://huggingface.co/google/gemma-2-9b-it) හි access request කරන්න.
## 🔌 API Usage
```python
from gradio_client import Client
client = Client("your-username/srt-cleaner")
# Text input API
result = client.predict(
text_content="1\n00:00:01,000 --> 00:00:04,000\nwww.subscene.com\n\n2\n...",
use_gemma=True,
api_name="/handle_text_input"
)
cleaned_srt, report, output_file = result
```
## 📋 What Gets Removed?
- Website URLs (http://, www.)
- Subtitle site names (Subscene, OpenSubtitles, etc.)
- "Subtitles by / Translated by / Synced by" credits
- Copyright notices (©, All rights reserved)
- Promotional content detected by Gemma AI