KanoAI (ગુજરાતી KanoAI Suite)

The Comprehensive All-in-One Gujarati Language & AI Intelligence Suite.

An open-source ecosystem bridging classical Gujarati typography, dynamic stroke animations, real-time handwriting recognition, native audio generation, Speech-to-Text (STT), Text-to-Speech (TTS), and AI grammar intelligence.

🌐 Live Interactive Apps:

🖋️ Kano Stroke Animator & Audio: https://gajjartejas.github.io/KanoAI/
✍️ Kano Handwriting Recognition & Practice Suite: https://gajjartejas.github.io/KanoAI/handwriting/
🗣️ Kano Gujarati Voice (TTS) Studio: https://gajjartejas.github.io/KanoAI/#tts
📊 Kano Gujarati TTS Model Comparison (3-Column Matrix): https://gajjartejas.github.io/KanoAI/tts-comparison/
📸 Kano Gujarati OCR & Vision Studio: https://gajjartejas.github.io/KanoAI/#ocr

🖋️ Stroke Animator & Kano Audio Suite	✍️ Handwriting Practice Suite	🗣️ Gujarati Text-to-Speech (TTS) Studio	📸 Gujarati OCR & Vision Studio

⚡ Quickstart: Launch All Servers & Open Webpage

Run the one-click launcher to start the Python Unified Backend API Server (OCR + TTS on port 8000) and the Local Web Studio (port 8085), perform health checks, and automatically open the suite in your default browser:

# Using Bash (macOS / Linux):
./start_all.sh

# Or using Python (macOS / Windows / Linux):
python3 start_all.py

Options & Flags:

./start_all.sh --tab tts • Open directly to the Text-to-Speech (TTS) Studio (#tts)
./start_all.sh --tab ocr • Open directly to the OCR & Vision Studio (#ocr)
./start_all.sh --tab handwriting • Open directly to the Handwriting Suite (/handwriting/)
./start_all.sh --no-browser • Start servers in background without opening browser
./start_all.sh --api-port 8000 --web-port 8085 • Custom ports

🚀 The KanoAI Ecosystem Roadmap

Module	Status	Description
🖋️ Kano Trace	✅ Live	Stroke-by-stroke animation, EDT centerline extraction, and 565-character stroke catalog.
✍️ Kano Handwriting	✅ Live	Real-time offline recognition combining Sakoe-Chiba DTW + in-memory Tiny CNN.
🔊 Kano Audio	✅ Live	Compressed, crystal-clear native speech pronunciations for all 565 characters.
🗣️ Kano Voice (TTS)	✅ Live	State-of-the-art neural TTS with AI4Bharat IndicF5 & Indic-TTS / Bodhan Indic-Speak.
📸 Kano Vision (OCR)	✅ Live	Multi-engine Gujarati OCR (Bhashini IndicPhotoOCR, Gujarati TrOCR, GujaratiHCR) with voice playback bridge.
🎙️ Kano Listen (STT)	📋 Planned	Offline & low-latency Gujarati Speech-to-Text acoustic modeling.
🧠 Kano Grammar (AI)	📋 Planned	LLM-assisted Gujarati spell-checker, grammar analysis, sandhi/samasa parser, and NLP toolkits.
⚡ Kano API	✅ Live	Lightweight REST / JSON microservices for characters, strokes, phonemes, and audio synthesis.

📁 Repository Structure

KanoAI/
├── .agents/                      # Antigravity AI agent skills & custom workflows (/update-markdown)
├── handwriting/                  # Real-Time Gujarati Handwriting Recognition & Practice Workspace
│   ├── src/                      # Canvas, Guided Tracing, Quiz, Free Draw, Hybrid DTW + Tiny CNN
│   ├── __tests__/                # Automated test suites (30 tests)
│   ├── scripts/                  # Template extraction & model generation scripts
│   ├── package.json              # TypeScript, React Native Web, Expo, and Jest runners
│   └── README.md                 # Full technical spec & architecture documentation
├── node/                         # Node.js resource generation workspace
│   ├── package.json              # Dependencies (text-to-svg, node-fetch)
│   ├── index.js                  # Static SVG, CSV, and TTS audio generator
│   └── README.md                 # Node.js documentation and usage instructions
├── python/                       # Python workspace
│   ├── setup_venv.sh             # Industry-standard virtual environment setup script
│   ├── README.md                 # Python workspace overview & venv instructions
│   ├── ocr/                      # Multi-Engine Gujarati OCR Studio (IndicPhotoOCR, TrOCR, GujaratiHCR)
│   ├── tts/                      # Neural Gujarati Voice Studio (IndicF5 & Indic-TTS)
│   ├── tests/                    # Comprehensive unit tests for OCR, TTS, APIs & services
│   └── char_stroke_generation/   # Dedicated character stroke generation package
│       ├── stroke_generator/     # Core library (Bézier, EDT, HarfBuzz, IoU matching)
│       ├── scripts/              # Standalone CLI runners
│       ├── tests/                # Automated unit tests
│       ├── pyproject.toml        # PEP 621 packaging
│       ├── requirements.txt      # Pip dependencies
│       └── README.md             # Technical documentation & math formulations
├── docs/                         # Live GitHub Pages interactive frontend & Kano audio suite
│   ├── index.html                # Semantic responsive single-page web application
│   ├── css/                      # Modular styling (main, stage comparison, ocr, character grid)
│   ├── js/                       # Stage renderer, Kano audio player, OCR studio, app coordinator
│   ├── assets/                   # Full 565-character SVG catalog, OCR sample documents & previews
│   └── handwriting/              # Production web build for the Handwriting Recognition Suite
├── scripts/                      # Launchers (start_all.sh/py), test runners (run_tests.sh/py) & asset tools
├── start_all.sh                  # One-click suite launcher (Bash)
├── start_all.py                  # One-click suite launcher (Python)
├── fonts/                        # Shared TrueType/OpenType Gujarati fonts
├── resources/                    # Shared JSON definitions & raw datasets (Kakko, Barakhadi, Numerals)
├── interpolate-svg/              # Manual reference SVG stroke templates
├── output/                       # Generated SVG artifacts & preview catalogs (.gitignored)
└── .github/workflows/            # GitHub Actions automated GitHub Pages deployment workflow

🌐 Live Interactive Frontend & Kano Audio Suite (`docs/`)

🔗 Live Production Demo: https://gajjartejas.github.io/KanoAI/

The live web application provides an interactive stroke animator and audio player:

All 565 Gujarati Characters: Complete Kakko (45), full Barakhadi (420 across 35 consonants), and Numerals 0–100 (101).
Normalized 1:1 Stage Sizing: Explicit viewBox coordinate normalization ensuring Reference (Light) and Auto-Generated (Bold) comparison cards scale with equal proportions and perfect centering.
Dedicated Settings Panel: Right-side desktop dock with quick playback controls, layer toggles, speed adjustments, and keyboard shortcuts.
Dynamic Stroke Width Options: Interactive slider (1px to 10px) with live preset pills (2px, 3.5px, 5px, 7px) modifying both stages in real-time.
Kano Audio Pronunciations: Speech pronunciations for all 565 characters, compressed via ffmpeg to high-efficiency MP3s, featuring auto-play on selection, dedicated 🔊 Pronounce button, and persistent Sound: ON/OFF toggle.
Dark / Light Theme Toggle: Top-right switch with Sun/Moon icons, custom HSL color palettes, and localStorage memory.
Mobile Responsive Design: Clean side-by-side stages, compact slider grids, and touch-optimized character targets on mobile viewports.
Zero-Click GitHub Pages Deployment: Fully automated via .github/workflows/deploy-pages.yml.

🖥️ Stroke Animator & Audio Previews

Dark Theme (Default)	Light Theme

✍️ Real-Time Gujarati Handwriting Recognition & Practice Suite (`handwriting/`)

🔗 Live Handwriting Web App: https://gajjartejas.github.io/KanoAI/handwriting/

A complete handwriting recognition and practice suite featuring:

Guided Practice: Step-by-step Gujarati character tracing with directional hints, sequential stroke bubbles, and instant feedback.
Animated Player: Interactive stroke-by-stroke playback of all 565 characters with speed controls and printable worksheet generator.
Quiz Game: Gamified learning pairing Gujarati vocabulary, native audio pronunciation, and interactive drawing challenges.
Free Drawing & Recognition: Multi-stroke canvas powered by a Hybrid Recognition Engine combining Sakoe-Chiba Dynamic Time Warping (DTW) and an in-memory Tiny CNN (Conv2D/MaxPool/Dense) classifier.
Accuracy Benchmarks: Automated performance testing verifying sub-20ms latency and high recognition accuracy across characters.

🖥️ Handwriting Suite Preview

cd handwriting
npm install
npm test            # Run all 7 Jest test suites (30 tests)
npm run web         # Launch local development server
npm run build:web   # Export production bundle to docs/handwriting

👉 See Handwriting Documentation (handwriting/README.md) for architecture details.

🗣️ Gujarati Text-to-Speech (TTS) Studio (`python/tts/`)

State-of-the-Art Gujarati Neural Voice Generation integrating 100% offline local models and AI4Bharat architectures:

Meta MMS-TTS (facebook/mms-tts-guj) ★ 100% Offline & Local:
- Architecture: End-to-end Variational Inference with Adversarial Learning (VITS, 16,000 Hz).
- Execution: 100% offline and local inference on Apple Silicon MPS, NVIDIA CUDA, or CPU with zero cloud rate limits.
- Capabilities: High-fidelity isolated consonant, syllable, Barakhadi, and numeral synthesis with automated verbalization (૧ → એક, ૧૦ → દસ).
Piper TTS (piper-tts) ★ Ultra-Fast ONNX Runtime:
- Architecture: Neural text-to-speech running locally on ONNX Runtime with sub-100ms inference on low-resource CPUs.
- Models: Native Hindi / Indic voices (Rohan, Pratham, Priyamvada via rhasspy/piper-voices) and community Gujarati checkpoint (Arjun4707/piper-gujarati-male trained on AI4Bharat dataset).
- Capabilities: Ultra-lightweight CPU execution, numeral verbalization, and seamless cross-venv bridge.
Parler-TTS Indic (ai4bharat/indic-parler-tts) ★ Prompt-Guided Expressive Audio:
- Architecture: Natural language audio codec synthesis (24,000 Hz) guided by descriptive English prompts.
- Presets: Warm & Gentle Female, Expressive Storyteller, Authoritative Broadcaster Male, Intimate Whisper, or custom freeform prompts.
- Endpoints: Local Engine (port 7861), Cloud HuggingFace Spaces ZeroGPU, or Custom Endpoint URL.
Browser Native Web Speech API ★ Zero-Latency Client-Side:
- Architecture: W3C SpeechSynthesis hardware-accelerated synthesis executing directly inside the client's browser.
- Capabilities: Zero server latency (0ms), dynamic pitch/rate controls, automatically detects OS Gujarati (gu-IN) voices (Apple & Google).
Bhashini Official Neural TTS (Digital India / IIT Madras) ★ National Benchmark:
- Architecture: Official Digital India Bhashini Mission benchmark neural models for formal academic and news diction.
- Voice Profiles: Bhashini Official Female (Priyamvada neural model fallback) and Bhashini Official Male.
- Endpoints: 100% Local Offline Engine (no API key needed), Official ULCA Cloud API (dhruva-api.bhashini.gov.in), or Custom URL.
Google Cloud Neural2 / Wavenet (gu-IN) ★ Studio Quality Commercial Neural:
- Architecture: Google DeepMind Wavenet, Standard, and Neural2 models (gu-IN-Wavenet-A/B, gu-IN-Neural2-A/B).
- Endpoints: Official Google Cloud REST API with Linear16 audio, Local Offline Neural Fallback, or Custom Proxy URL.
AI4Bharat IndicF5 (GitHub / HuggingFace):
- Architecture: Flow-matching diffusion-style voice synthesis (24,000 Hz).
- Endpoints: Local Server (port 7865), Cloud Hugging Face Space, or Custom Endpoint URL.
AI4Bharat Indic-TTS / Bodhan-AI Indic-Speak (GitHub / HuggingFace):
- Architecture: Multi-speaker fast neural acoustic model (44,100 Hz).
- Endpoints: Local Server (port 7861), Cloud Hugging Face Space, or Custom Endpoint URL.
eSpeak-NG (Python Wrapper & Formant Engine) ★ Lightweight Phoneme Debugger:
- Architecture: Formant-based synthesis, lightweight (<10 MB), zero neural weights.
- Languages: Native gu (Gujarati) and hi (Hindi) phoneme tables.
- Capabilities: Instant offline phonetic IPA and consonant debugging (swars, vyanjans, matras, virama) with pure acoustic formant synthesis.

🖥️ TTS Studio Preview

🚀 Running the TTS API Server & Web Studio

# 1. Start the lightweight TTS HTTP API server (port 8000)
.venv/bin/python3 python/tts/server.py --port 8000

# 2. Open the web studio in docs/ or via local server
python3 -m http.server 8085 --directory docs
# Navigate to: http://localhost:8085/#tts

🎙️ Standalone Batch Audio Generator (`python/audio_generation/`)

Dedicated offline batch pronunciation generator for Kakko, Barakhadi series, and numerals with optional FFmpeg MP3 compression:

# 1. Run pronunciation benchmark across sample isolated characters & numerals:
.venv/bin/python3 python/audio_generation/generate_tts_local.py --benchmark

# 2. Synthesize custom Gujarati text directly to MP3:
.venv/bin/python3 python/audio_generation/generate_tts_local.py --text "કક્કો" --output out.mp3 --format mp3

# 3. Batch generate isolated Kakko consonants (WAV or MP3):
.venv/bin/python3 python/audio_generation/generate_tts_local.py --category kakko --output-dir output/tts/kakko --format mp3

# 4. Batch generate numerals (૧ to ૧૦૦):
.venv/bin/python3 python/audio_generation/generate_tts_local.py --category numerals --output-dir output/tts/numerals --format mp3

💻 TTS Command Line Interface (CLI)

# Synthesize with Meta MMS-TTS (Offline VITS)
.venv/bin/python3 python/tts/cli.py --engine mms_tts --text "નમસ્તે KanoAI" --output output/tts/mms.wav

# Synthesize with Piper TTS (Rohan - ONNX Runtime)
.venv/bin/python3 python/tts/cli.py --engine piper_tts --voice rohan --text "નમસ્તે KanoAI" --output output/tts/piper.wav

# Synthesize and debug phonemes with eSpeak-NG (Gujarati)
.venv/bin/python3 python/tts/cli.py --engine espeak_ng --lang gu --text "નમસ્તે KanoAI" --output output/tts/espeak.wav

# Synthesize with Indic-TTS (Dhara)
.venv/bin/python3 python/tts/cli.py --engine indic_tts --text "નમસ્તે KanoAI" --output output/tts/dhara.wav

# Synthesize with IndicF5
.venv/bin/python3 python/tts/cli.py --engine indic_f5 --text "નમસ્તે KanoAI" --output output/tts/f5.wav

# Multi-engine side-by-side comparison across all 5 engines
.venv/bin/python3 python/tts/cli.py --compare --text "ગુજરાતી ભાષા ખૂબ જ સુંદર છે."

# Run automated test suite
.venv/bin/python3 python/tts/cli.py --test

📸 Gujarati OCR & Vision Studio (`python/ocr/`)

Multi-Engine Gujarati Optical Character Recognition (OCR) covering printed literature, official scanned certificates, camera scene text / signboards, and handwritten notebook pages.

Bhashini-IITJ IndicPhotoOCR (GitHub / HuggingFace Space):
- Architecture: DBNet Text Detector + Indic Multilingual Sequence Recognition.
- Best For: Natural scene text, street signs, posters, and full page document localization with bounding box coordinates.
Gujarati TrOCR (HuggingFace):
- Architecture: Vision Transformer (TrOCR) encoder-decoder trained on Gujarati typography and conjuncts.
- Best For: Scanned books and printed literature (96.2% exact word accuracy, 1.35% CER).
GujaratiHCR Research Pipeline:
- Architecture: Adaptive binarization, horizontal/vertical projection profiling for line/word segmentation, contour isolation, and CNN-LSTM recognition.
- Best For: Cursive Gujarati handwriting, ruled notebook pages, and individual handwritten characters.

🖥️ OCR Studio Preview

🚀 Running the Unified OCR & TTS Backend Server

# 1. Start the unified OCR & TTS HTTP API server (port 8000)
.venv/bin/python3 python/ocr/server.py --port 8000

# 2. Open the web studio in docs/ or via local server
python3 -m http.server 8085 --directory docs
# Navigate to: http://localhost:8085/#ocr

💻 OCR Command Line Interface (CLI)

# Recognize scene photo or document with Bhashini IndicPhotoOCR
.venv/bin/python3 python/ocr/cli.py --image docs/assets/ocr_samples/sample_1_printed_book.png --engine indic_photo_ocr

# Recognize printed document with Gujarati TrOCR
.venv/bin/python3 python/ocr/cli.py --image docs/assets/ocr_samples/sample_4_official_doc.png --engine gujarati_trocr

# Run Handwriting HCR line segmentation
.venv/bin/python3 python/ocr/cli.py --image docs/assets/ocr_samples/sample_3_handwritten_note.png --engine gujarati_hcr

# Compare all 3 OCR engines side-by-side
.venv/bin/python3 python/ocr/cli.py --image docs/assets/ocr_samples/sample_1_printed_book.png --compare

# Run OCR automated self-test
.venv/bin/python3 python/ocr/cli.py --test

👉 See OCR Documentation (python/ocr/README.md) for full API specifications.

🧪 Automated Unit Testing & CI Verification (`python/tests/`)

The KanoAI suite includes a comprehensive Python unit test suite verifying OCR, TTS, API request handlers, image segmentation, and system launchers:

🚀 Running All Tests

# Run using the automated shell test runner (macOS / Linux):
./scripts/run_tests.sh

# Or run using the cross-platform Python runner:
python3 scripts/run_tests.py

# Or directly with Python standard library unittest:
PYTHONPATH=python python3 -m unittest discover -s python/tests -p "test_*.py" -v

📋 Test Coverage Matrix (37 Tests)

Component	Test Module	Description
📸 OCR Service Coordinator	`test_ocr_service.py`	Multi-engine routing (`indic_photo_ocr`, `gujarati_trocr`, `gujarati_hcr`), automatic fallback chain, sample catalog schema, comparison mode.
✍️ GujaratiHCR Segmentation	`test_ocr_hcr.py`	Grayscale & Otsu binarization, horizontal/vertical projection slicing, normalized bounding box coordinates (`norm_x`, `norm_y`, `norm_w`, `norm_h`), network timeout fallback.
📖 Gujarati TrOCR	`test_ocr_trocr.py`	Multi-line document contour slicing, input formats (PIL, NumPy, base64 data URIs, files), Hugging Face router & local server mocks.
⚡ OCR Server API	`test_ocr_server.py`	HTTP microservice endpoints (`/api/health`, `/api/ocr/samples`, `/api/ocr/recognize`, `/api/ocr/compare`), CORS preflight OPTIONS, 400 Bad Request handling.
🗣️ TTS Service Coordinator	`test_tts_service.py`	Unified preset catalogue, dual-engine voice routing (`indic_f5`, `indic_tts`), side-by-side comparison synthesis, speaker registry.
🎙️ Neural TTS Engines	`test_tts_engines.py`	`IndicF5Engine` and `IndicTTSEngine` parameter validations, empty input assertions, base64 audio payload encoding, speaker profiles (`Dhara`, `Parth`).
⚡ TTS Server API	`test_tts_server.py`	HTTP microservice endpoints (`/api/health`, `/api/tts/voices`, `/api/tts/synthesize`, `/api/tts/presets`), CORS headers, malformed payload recovery.
🚀 System Launchers	`test_launcher.py`	Cross-platform port availability check (`is_port_in_use`), health check wait loop (`wait_for_url`), repository root path resolution.

🟢 Node.js Workspace (`node/`)

Used for rendering standard glyph SVGs, CSV matrices, and Google Wavenet audio files.

cd node
npm install
node index.js

👉 See Node.js Documentation (node/README.md) for details.

🐍 Python Workspace: Character Stroke Generation (`python/`)

An analytical geometry and Euclidean Distance Transform (EDT) pipeline to automatically extract medial centerline strokes from Gujarati fonts for the Kano React Native educational app.

Quick Start (Industry-Standard Virtual Environment Setup)

# 1. Run automated environment setup (creates .venv & installs dependencies)
./python/setup_venv.sh

# 2. Activate virtual environment
source .venv/bin/activate

# 3. Run automated unit tests
PYTHONPATH=python/char_stroke_generation python -m unittest discover -s python/char_stroke_generation/tests -t python/char_stroke_generation

# 4. Generate 56 showcase samples and launch interactive viewer
python python/char_stroke_generation/scripts/build_samples.py
python python/char_stroke_generation/scripts/build_viewer.py --serve 8765

👉 See Python Workspace (python/README.md) and Stroke Generation Guide (python/char_stroke_generation/README.md) for details.

📚 Character Dataset & Catalog

The KanoAI suite catalogs and interactively supports 565 total characters:

Kakko (45): Complete Gujarati vowels (સ્વર) and consonants (વ્યંજન).
Barakhadi (420): Comprehensive matra combinations across 35 consonants.
Numerals (101): Gujarati digits and numbers ૦ to ૧૦૦ (0–100) with complete transliterations and names.

Full structured JSON datasets and definitions are located in resources/:

All characters can be interactively browsed, animated, pronounced, and practiced in the Live Suite.

🗺️ Roadmap & Upcoming Milestones

Kano Trace: 565-character stroke animation & analytical EDT centerline extraction.
Kano Handwriting: Dual DTW + Tiny CNN offline recognition engine (<20ms latency).
Kano Voice (TTS): Neural Gujarati speech synthesis model (AI4Bharat IndicF5 & Indic-TTS / Bodhan-AI Indic-Speak) with smooth 60fps waveform sync.
Kano Vision (OCR): Multi-engine Gujarati OCR studio (Bhashini-IITJ IndicPhotoOCR, Gujarati TrOCR, GujaratiHCR) with voice playback bridge.
Kano Listen (STT): Offline Speech-to-Text engine optimized for regional accents.
Kano Grammar AI: Contextual spell-checker, Sandhi/Samasa decomposition, and morphological analysis.
Kano Cloud API: Developer REST/GraphQL endpoints for character stroke vectors, phonetics, and datasets.

🙏 Third-Party Credits & Acknowledgements

KanoAI proudly stands on the shoulders of the open-source and scientific research communities. We gratefully acknowledge the following projects, models, datasets, and libraries:

🧠 Neural AI Models & Machine Learning Research

Meta MMS-TTS (Meta AI Research): Scaling Speech Technology to 1,000+ languages; Variational Inference with Adversarial Learning (VITS) Gujarati neural speech checkpoint (facebook/mms-tts-guj).
Piper TTS (Rhasspy / Michael Hansen & community): Fast, local neural text-to-speech engine running on ONNX Runtime (Hugging Face Voices).
Arjun4707/piper-gujarati-male: Community Gujarati Piper model trained on the AI4Bharat dataset.
Digital India Bhashini Mission / IIT Madras: National Language Translation Mission (NLTM) official benchmark neural text-to-speech and Dhruva ULCA inference pipelines.
Google Cloud Text-to-Speech: Google DeepMind WaveNet & Neural2 commercial neural speech synthesis models for Gujarati (gu-IN).
AI4Bharat Parler-TTS Indic (IIT Madras & AI4Bharat): Natural language prompt-guided expressive text-to-speech for Indian languages.
AI4Bharat IndicF5 (IIT Madras & AI4Bharat): Reference-conditioned Flow-Matching Speech Synthesis for Indian languages (Hugging Face Space).
AI4Bharat Indic-TTS (IIT Madras & AI4Bharat): Multi-speaker neural text-to-speech acoustic models for Indian languages (Hugging Face).
Bodhan AI (Indic-Speak): Natural Gujarati female (Dhara) and male (Parth) neural speaker profiles.
eSpeak-NG: Multi-lingual software speech synthesizer with zero neural weights, formant acoustic modeling, and native Gujarati (gu) & Hindi (hi) phoneme tables.
W3C Web Speech API: Client-side device-native hardware speech synthesis specification for zero-latency in-browser vocalization.
Bhashini-IITJ IndicPhotoOCR (Digital India Bhashini & IIT Jodhpur): Scene text detection (DBNet) and multilingual Indic sequence recognition (Hugging Face Space).
Gujarati TrOCR by Umang Chaudhari: Vision Transformer (ViT + RoBERTa) encoder-decoder fine-tuned specifically for Gujarati text and conjuncts.
Microsoft TrOCR (Microsoft Research): Transformer-based Optical Character Recognition architecture for document processing.

🔤 Open-Source Fonts & Typography

Google Noto Fonts: Noto Sans Gujarati and Noto Serif Gujarati by Google Fonts and the Monotype design team (SIL Open Font License).
Gujarati Open-Source Typefaces: Rasa, Mogra, Shrikhand, Padauk, and Anek Gujarati created and published under the SIL Open Font License (OFL).

🛠️ Core Libraries & Open-Source Software

OpenCV (cv2): Real-time computer vision, Otsu adaptive thresholding, morphological filtering, and contour extraction.
PyTorch & Hugging Face Transformers: Tensor processing, tokenizer runtime, and neural inference pipelines.
Gradio Client: Asynchronous event-driven streaming client for neural inference endpoints.
Pillow (PIL): Raster image processing and format decoding.
SoundFile & NumPy: High-performance numerical computing and WAV audio serialization.
React Native Web & Expo: Multi-platform web application runtime for the handwriting practice suite.
Jest: Fast and delighting JavaScript testing framework.
Simple Icons & Lucide Icons: High-quality SVG brand icons and clean UI vectors.

📄 License

KanoAI is licensed under the GNU GENERAL PUBLIC LICENSE.

Navigation Menu

Repository navigation

About

Repository files navigation

KanoAI (ગુજરાતી KanoAI Suite)

⚡ Quickstart: Launch All Servers & Open Webpage

🚀 The KanoAI Ecosystem Roadmap

📁 Repository Structure

🌐 Live Interactive Frontend & Kano Audio Suite (`docs/`)

🖥️ Stroke Animator & Audio Previews

✍️ Real-Time Gujarati Handwriting Recognition & Practice Suite (`handwriting/`)

🖥️ Handwriting Suite Preview

🗣️ Gujarati Text-to-Speech (TTS) Studio (`python/tts/`)

🖥️ TTS Studio Preview

🚀 Running the TTS API Server & Web Studio

🎙️ Standalone Batch Audio Generator (`python/audio_generation/`)

💻 TTS Command Line Interface (CLI)

📸 Gujarati OCR & Vision Studio (`python/ocr/`)

🖥️ OCR Studio Preview

🚀 Running the Unified OCR & TTS Backend Server

💻 OCR Command Line Interface (CLI)

🧪 Automated Unit Testing & CI Verification (`python/tests/`)

🚀 Running All Tests

📋 Test Coverage Matrix (37 Tests)

🟢 Node.js Workspace (`node/`)

🐍 Python Workspace: Character Stroke Generation (`python/`)

Quick Start (Industry-Standard Virtual Environment Setup)

📚 Character Dataset & Catalog

🗺️ Roadmap & Upcoming Milestones

🙏 Third-Party Credits & Acknowledgements

🧠 Neural AI Models & Machine Learning Research

🔤 Open-Source Fonts & Typography

🛠️ Core Libraries & Open-Source Software

📄 License

About

Releases

Packages

Contributors

Languages

Name		Name	Last commit message	Last commit date
.agents/skills/update-markdown		.agents/skills/update-markdown
.github/workflows		.github/workflows
.vscode		.vscode
docs		docs
fonts/Noto_Sans_Gujarati		fonts/Noto_Sans_Gujarati
handwriting		handwriting
interpolate-svg/svgs		interpolate-svg/svgs
node		node
python		python
resources		resources
scripts		scripts
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
STROKE_GENERATION_GUIDE.md		STROKE_GENERATION_GUIDE.md
requirements.txt		requirements.txt
start_all.py		start_all.py
start_all.sh		start_all.sh
viewer.html		viewer.html

Latest commit

History

Folders and files

Repository files navigation

KanoAI (ગુજરાતી KanoAI Suite)

⚡ Quickstart: Launch All Servers & Open Webpage

🚀 The KanoAI Ecosystem Roadmap

📁 Repository Structure

🌐 Live Interactive Frontend & Kano Audio Suite (docs/)

🖥️ Stroke Animator & Audio Previews

✍️ Real-Time Gujarati Handwriting Recognition & Practice Suite (handwriting/)

🖥️ Handwriting Suite Preview

🗣️ Gujarati Text-to-Speech (TTS) Studio (python/tts/)

🖥️ TTS Studio Preview

🚀 Running the TTS API Server & Web Studio

🎙️ Standalone Batch Audio Generator (python/audio_generation/)

💻 TTS Command Line Interface (CLI)

📸 Gujarati OCR & Vision Studio (python/ocr/)

🖥️ OCR Studio Preview

🚀 Running the Unified OCR & TTS Backend Server

💻 OCR Command Line Interface (CLI)

🧪 Automated Unit Testing & CI Verification (python/tests/)

🚀 Running All Tests

📋 Test Coverage Matrix (37 Tests)

🟢 Node.js Workspace (node/)

🐍 Python Workspace: Character Stroke Generation (python/)

Quick Start (Industry-Standard Virtual Environment Setup)

📚 Character Dataset & Catalog

🗺️ Roadmap & Upcoming Milestones

🙏 Third-Party Credits & Acknowledgements

🧠 Neural AI Models & Machine Learning Research

🔤 Open-Source Fonts & Typography

🛠️ Core Libraries & Open-Source Software

📄 License

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

🌐 Live Interactive Frontend & Kano Audio Suite (`docs/`)

✍️ Real-Time Gujarati Handwriting Recognition & Practice Suite (`handwriting/`)

🗣️ Gujarati Text-to-Speech (TTS) Studio (`python/tts/`)

🎙️ Standalone Batch Audio Generator (`python/audio_generation/`)

📸 Gujarati OCR & Vision Studio (`python/ocr/`)

🧪 Automated Unit Testing & CI Verification (`python/tests/`)

🟢 Node.js Workspace (`node/`)

🐍 Python Workspace: Character Stroke Generation (`python/`)