Sanskrit Karaoke
Created: 12-05-2026 · Last updated: 10-10-2026
Wave-notation visualiser and karaoke exporter for Sanskrit verse
Live app → · v1.5.5 · changelog
If the app doesn't reflect the latest version after an update, do a hard refresh to clear the cache:
Edge, Firefox, Opera — Ctrl+F5 · Chrome — Ctrl+Shift+R · Safari (Mac) — Cmd+Shift+R or Cmd+Option+R
Sanskrit Karaoke turns a Sanskrit śloka into an interactive wave diagram that shows the metrical weight of each syllable (guru / laghu), lets you mark audio timing, and exports karaoke videos or high-resolution images.
Features
- Wave diagram — each syllable appears as a circle on one of five horizontal rows; heavy syllables (guru) in dark red, light ones (laghu) in dark green; drag to rearrange
- Encoding support — paste text in Devanagari, IAST, SLP1, Harvard-Kyoto, ITRANS, Velthuis, or WX; encoding is auto-detected
- Meter detection — automatically identifies the meter (anuṣṭubh, samavṛtta, jāti, and others) and marks vipulā variants in anuṣṭubh padas
- Audio timing editor — two modes:
- Pada mode: drag eight boundary lines to mark the start/end of each pada
- Syllable mode: fine-tune timing per syllable with keyboard shortcuts
- Karaoke MP4 export — renders the wave diagram with the audio track and an animated highlight dot
- Sidecar captions —
node tools/export_captions.mjs <verse.json> [--timing timing.json] writes .srt/.vtt from syllable timing without rendering video
- PNG export — 1920 × 1080 image with Devanagari, IAST transliteration, and footer metadata
- Google Drive — save and load sessions (JSON + audio) to a shared Drive folder
- Mobile & touch — full touch drag in the Timing Editor; on-screen navigation bar (◀ ▶ −0.01 +0.01 ⌂) on touch devices; iOS-specific export overlays for PNG and MP4; iPhone safe-area support
Using the app
- Open the live app.
- Paste the first and second half-verse (s1 / s2) into the text fields.
- Press Run — the wave diagram appears and the meter is identified.
- Adjust syllable weights or rows by right-clicking any syllable.
- Load an audio file and open the Timing Editor:
- Use Pada mode first to mark rough boundaries.
- Switch to Syllable mode for precise per-syllable timing.
- Export with PNG or Karaoke MP4.
Example inputs
Paste s1 into the first field, s2 into the second. Encoding is auto-detected (Devanagari, IAST, SLP1, etc.).
Anuṣṭubh (8 syllables × 4 pādas) — Bhagavadgītā 2.47
| Field | Text |
|---|
| s1 | कर्मण्येवाधिकारस्ते मा फलेषु कदाचन |
| s2 | मा कर्मफलहेतुर्भूर्मा ते सङ्गोऽस्त्वकर्मणि |
Upajāti / Triṣṭubh (11 syllables × 4 pādas) — Kumārasambhava 1.1 (Kālidāsa)
| Field | Text |
|---|
| s1 | अस्त्युत्तरस्यां दिशि देवतात्मा हिमालयो नाम नगाधिराजः |
| s2 | पूर्वापरौ तोयनिधी वगाह्य स्थितः पृथिव्या इव मानदण्डः |
Mandākrāntā (17 syllables × 4 pādas) — Meghadūta 1.1 (Kālidāsa)
Each printed line of Meghadūta is one pāda. Concatenate lines 1+2 into s1 and lines 3+4 into s2.
| Field | Text |
|---|
| s1 | कश्चित्कान्ताविरहगुरुणा स्वाधिकारात्प्रमत्तः शापेनास्तङ्गमितमहिमा वर्षभोग्येण भर्तुः |
| s2 | यक्षश्चक्रे जनकतनयास्नानपुण्योदकेषु स्निग्धच्छायातरुषु वसतिं रामगिर्याश्रमेषु |
Keyboard shortcuts (Timing Editor — syllable mode)
| Key | Action |
|---|
← / → | Select previous / next syllable |
Ctrl+← / Ctrl+→ | Shift syllable timing −0.01 s / +0.01 s |
Home | Align syllable to pada start |
Enter | Play current syllable |
Space | Play current syllable and advance |
Ctrl+Space | Play to end of śloka |
Ctrl+Enter | Play to end of pada |
Running locally
cd SanskritKaraoke
python -m http.server 8000
# open http://localhost:8000
No build step or package manager required.
Student page generation:
python tools/make_student.py
python tools/make_student.py --check
tools/templates/student.html is the source of truth for the deployed student.html.
Syntax check:
node --check src/scripts/app.js
Code overview
| File | Contents |
|---|
index.html | Entire UI — main view + all modals (settings, help, timing editor, Drive picker) |
src/scripts/app.js | Main application logic (~320 KB ES module) — imports the reusable rendering core |
src/core/*.js | Extracted rendering core (translit, layout, svg, compose, karaoke-frame, timing, feed) shared by the app and the headless renderer — see ADR-0001 |
docs/reference/apte_prosody.html | Apte prosody reference database |
ARCHITECTURE.md | Current + target architecture, phase by phase |
Teaching Pipeline & Evals
The project includes a server-side Teaching Pipeline built on LangGraph 1.0 and Pydantic, designed to automate verse curation and student analysis.
Architecture
- VerseCurator: Validates and enriches new verse JSON files.
- ContentEnricher: Uses Gemini Flash to generate missing translations and tags.
- QualityGate: Enforces strict semantic and schema rules before publication.
- StudentAnalyzer: Analyzes SRS history and recommends the next study queue.
This is the beginning of the README. Read the rest at the original. Read the full README ↗
