Update README.md

README.md

@@ -1,12 +1,57 @@
 ---
-title: Translator
-emoji: 🐠
-colorFrom: yellow
-colorTo: pink
-sdk: static
+title: Urdu Hindi English AI Translator
+emoji: 🌐
+colorFrom: blue
+colorTo: indigo
+sdk: gradio
+sdk_version: 5.49.1
+app_file: app.py
 pinned: false
-license: odbl
-short_description: UDtoENG
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+license: apache-2.0
+python_version: "3.11"
+Urdu, Hindi, Roman Urdu and Hinglish → English Translator
+A Hugging Face Spaces-ready Gradio application for translating text, speech, video, documents, and scanned images into fluent English.
+Included features
+Urdu, Hindi, Roman Urdu, and Hinglish input
+Automatic language detection
+Faster-Whisper speech recognition
+Browser microphone and audio upload
+Video audio extraction using FFmpeg
+PDF, DOCX, TXT, RTF, and image extraction
+OCR for scanned pages and images
+Free local translation path using open-source models
+Optional OpenAI translation when `OPENAI_API_KEY` is configured
+Natural, professional, business, academic, simplified, summary, and key-point output modes
+Temporary session-only file handling
+Deploy on Hugging Face Spaces
+Create a new Hugging Face Space.
+Select Gradio as the SDK.
+For speech and local translation, choose a GPU hardware tier when available. CPU works but model loading and transcription will be slower.
+Upload every file and folder from this package to the root of the Space repository.
+Hugging Face will build and launch the application automatically.
+The repository root must contain:
+```text
+app.py
+README.md
+requirements.txt
+packages.txt
+translator_app/
+```
+Optional Space secrets
+In Settings → Variables and secrets, add any desired values:
+`OPENAI_API_KEY`: enables the OpenAI translation mode
+`OPENAI_MODEL`: defaults to `gpt-4.1-mini`
+`WHISPER_MODEL`: defaults to `small`; use `base` for low-memory CPU Spaces
+`NLLB_MODEL`: defaults to `facebook/nllb-200-distilled-600M`
+`FLAN_MODEL`: defaults to `google/flan-t5-base`
+`MAX_UPLOAD_MB`: defaults to `200`
+Never place API keys directly in source files.
+Hardware guidance
+The default models can require several gigabytes of RAM and download time on first use. A GPU Space is recommended for responsive speech transcription. For a smaller CPU deployment, set `WHISPER_MODEL=base` and reduce `MAX_UPLOAD_MB`.
+Privacy
+There is no account system or application database. Uploads are processed within the active Space container. Temporary job directories are deleted after processing, although platform infrastructure and logs remain subject to Hugging Face's own policies.
+Limitations
+Exact visual reconstruction of translated PDF and DOCX layouts is not included.
+Roman Urdu and Hinglish spelling is non-standard, so contextual quality varies in free local mode.
+Local model downloads may make the first request slower.
+Free Spaces may sleep when inactive and have limited memory or compute.