update README.md

README.md

@@ -35,31 +35,6 @@ Automatic forced alignment for Quran recitations. Upload an audio recording of a
 | [hetchyy/r7](https://huggingface.co/hetchyy/r7) | Phoneme ASR (Large — higher accuracy) |
 | [hetchyy/Quran-phoneme-mfa](https://huggingface.co/spaces/hetchyy/Quran-phoneme-mfa) | MFA forced alignment (external Space) |
 
-## Running locally
-
-```bash
-# Install dependencies
-pip install -r requirements.txt
-
-# Start the app (port 7860)
-python app.py
-
-# Dev mode — skip model preloading for fast startup
-python app.py --dev
-
-# With a public sharing link
-python app.py --share
-```
-
-### Optional: Cython acceleration
-
-The DP alignment inner loop has a Cython extension that provides 10-20x speedup. It is automatically compiled on startup, but if that fails (missing C compiler), the app falls back to pure Python.
-
-```bash
-# Manual build
-python setup.py build_ext --inplace
-```
-
 ## How it works
 
 ### Alignment algorithm
@@ -75,7 +50,7 @@ The core alignment uses **substring Levenshtein DP** with word-boundary constrai
 ### Retry and recovery
 
 When alignment fails for a segment:
-- **Tier 1:** Expanded search window (60 lookback, 40 lookahead)
+- **Tier 1:** Expanded search window
 - **Tier 2:** Expanded window + relaxed confidence threshold (0.45)
 - **Re-anchoring:** After 2 consecutive failures, n-gram voting re-localizes position within the surah
 
@@ -83,15 +58,4 @@ When alignment fails for a segment:
 
 Two playback modes with real-time word highlighting:
 - **Per-segment** — Animate a single aligned segment with word/character-level karaoke
-- **Mega card** — Unified text flow across all segments with click-to-seek and configurable opacity windowing (Reveal, Fade, Spotlight, Isolate, Consume modes)
-
-## Key dependencies
-
-- **[quranic-phonemizer](https://pypi.org/project/quranic-phonemizer/)** — Quran-specific grapheme-to-phoneme conversion with tajweed rules
-- **[recitations-segmenter](https://pypi.org/project/recitations-segmenter/)** — VAD model for Quran recitation audio
-- **torch 2.8** / **transformers 5.0** — Model inference
-- **Gradio ≥ 6.5.1** — Web UI framework
-
-## License
-
-MIT
+- **Mega card** — Unified text flow across all segments with click-to-seek and configurable opacity windowing (Reveal, Fade, Spotlight, Isolate, Consume modes)