Arivu Technologies
VOICE-FIRST AI APP - ARIVU PRODUCTV3 + V4 Released

NeoNote
Voice-First Intelligent Capture

Hold a button and speak. NeoNote transcribes raw, rambly speech into polished notes, automatically links mentions to contacts, extracts action items with due dates, and builds a visual force-directed knowledge graph of your personal memory web - all in real-time.

NEONOTE LIVE SPECACTIVE DEVELOPMENT
Platform VersionsV3 (Android/Kotlin) + V4 (Flutter)
Primary ASRGroq Cloud Whisper-Large-V3
Offline FallbackVosk + whisper.cpp JNI (V3)
Processing Latency<800ms draft - async cloud refine
UI Styles4 rendering modes
Color Palettes10 curated themes (Dark + Light)
Memory WebForce-directed knowledge graph
DB Tables5 SQLite tables (notes, contacts...)
01 / PRODUCT VISION

Your voice is your memory.
Talk to your day.

Manual note-taking, complex CRM data entry, and fragmented task management all require the same cognitive friction: you have to stop what you're doing to document it. NeoNote replaces that friction with one natural action - press and hold a button, speak.

Zero-Friction Capture

Press and hold to record. Release to process. No typing, no navigation required.

Speech-to-Clean-Thought

Raw, rambly speech stripped of filler words ('um', 'uh', 'like') and false starts - clean, readable output.

Automatic Contact Linking

Mentions of people automatically extracted and linked to the native contact timeline.

Proactive Action Items

Commitments and deadlines converted to follow-up tasks with extracted due dates and CalendarContract sync.

<800ms
Draft transcript on recording release
2 Platforms
Android native (V3) + Flutter cross-platform (V4)
10 Palettes
4 rendering styles × 10 curated themes
5 DB Tables
Notes, contacts, follow-ups, links, graph nodes
02 / DUAL-TIER PROCESSING ARCHITECTURE

On-device + cloud. Offline-capable. Always instant.

Free Tier / Offline

On-Device ASR

Vosk offline engine or whisper.cpp JNI binding - both run entirely on-device with no network dependency. Models downloaded once on first launch via background model manager.

Bundled Vosk speech recognition model
whisper.cpp JNI binding (V3 native)
Basic Regex NLP + local entity rule matching
Works with airplane mode on
Pro Tier / Cloud

Groq Cloud ASR + LLM

Groq Whisper-Large-V3 delivers average 300–700ms transcription latency for 30-second clips. LLM cleanup, entity extraction, and action item parsing run in the cloud pipeline.

Groq Whisper-Large-V3 (highest precision)
300–700ms average latency (30s clip)
LLM cleanup + entity extraction pipeline
Silent hot-swap when cloud result returns

Preview-Then-Refine Data Flow

Recording EndsUser releases button
Draft Preview <800msLocal engine produces raw transcript
Cloud ProcessingWAV sent to Groq async
Silent Hot-SwapClean note replaces draft
03 / NLP PIPELINE

4-stage text processing pipeline.

Every transcript runs through a 4-stage sequential pipeline: filler removal → summarization → contact matching → action item extraction. All stages run on-device in V3, in the cloud pipeline in V4.

01Text Cleaner
Regex patterns (Kotlin)

Strips disfluencies (um, uh, like, you know), removes repeated consecutive words (the the), eliminates false starts, capitalizes sentence beginnings. Zero latency - pure regex on-device.

02Summarizer
Rule-based + LLM (Pro)

Extracts a clean headline and 2–3 bullet key takeaways. Free tier uses extractive rule-based summarization. Pro tier uses Groq LLM for abstractive summaries and smart formatting.

03Entity Matcher
Fuzzy matching + NLP

Fuzzy-matches spoken names against the native Android contacts list (ContentResolver). Also extracts city and venue mentions for PLACE node population in the memory graph.

04Task Extractor
ActionItemExtractor.kt

Detects commitment triggers (need to, remind me to, call, send, schedule). Extracts relative date references (tomorrow, by Friday, next week) and computes epoch timestamps for CalendarContract sync.

04 / INSPECTABLE CODE

Real production code from NeoNote.

services/groq_transcription_service.dart

Uploads recorded PCM audio to Groq Cloud Whisper-Large-V3. On return, immediately runs client-side NLP: filler word removal, summary generation, entity extraction, and action item parsing.

1class GroqTranscriptionService {
2 static const String _groqUrl =
3 'https://api.groq.com/openai/v1/audio/transcriptions';
4
5 static Future<TranscriptionResult> transcribeAudio({
6 required String filePath,
7 required String apiKey,
8 }) async {
9 final request = http.MultipartRequest('POST', Uri.parse(_groqUrl))
10 ..headers['Authorization'] = 'Bearer $apiKey'
11 ..fields['model'] = 'whisper-large-v3'
12 ..files.add(await http.MultipartFile.fromPath('file', filePath));
13
14 final response = await request.send();
15 final body = await http.Response.fromStream(response);
16 final rawText = jsonDecode(body.body)['text'] ?? '';
17
18 return processTranscriptNLP(rawText); // Entity extraction + cleanup
19 }
20}
05 / KNOWLEDGE GRAPH & MEMORY WEB ENGINE

Every note. Every connection. One living graph.

Both V3 and V4 implement a bi-directional knowledge graph engine. Every note processed automatically updates entity nodes and weighted co-occurrence edges - creating a visual force-directed graph of personal memory.

Example: "Met Fahad in Dubai regarding API spec"
→ Node: "Fahad" [PERSON] - mention_count++
→ Node: "Dubai" [PLACE] - mention_count++
→ Edge: Fahad ↔ Dubai - weight: +1
→ Node: "API" [TAG] - mention_count++
PERSON Node

Glowing blue/violet bubbles with contact icons. Tap to view full contact timeline.

PLACE Node

Emerald green bubbles with location pin icons. Tap to see every note mentioning this place.

TAG Node

Warm gold pill badges. Topics, projects, concepts extracted from speech.

06 / DATABASE SCHEMA

5 SQLite tables. Full relational integrity.

TableKey ColumnsPurpose
notesid, raw_text, clean_text, summary, audio_path, duration_seconds, created_at, category_tag, mentioned_names, places_mentionedCore voice note records
contactsid, display_name, phone_number, email, last_mentioned, mention_countNative contact sync + mention tracking
follow_upsid, note_id (FK), text, is_completed, due_date, contact_idAction items with due date and contact link
note_contact_linksnote_id, contact_id (composite PK)Many-to-many: notes ↔ mentioned contacts
graph_nodesid, label, type (PERSON|PLACE|TAG), mention_count, last_updatedMemory web knowledge graph nodes
07 / V3 vs V4 COMPARISON

Two implementations. Same product.

FeatureV3 - Native Android (Kotlin)V4 - Flutter Cross-Platform
Technology StackKotlin + Jetpack Compose + CoroutinesDart + Flutter 3.x
Target PlatformsAndroid (API 24+ native)Android, iOS, Web
Database EngineNative SQLite (NeowispDbHelper)SQLite via sqflite (neonote_v4.db)
ASR EngineGroq Cloud + Local Vosk + whisper.cppGroq Cloud (whisper-large-v3)
Offline TranscriptionBundled Vosk / whisper.cpp JNICloud-primary with fallback stream
Device IntegrationFull ContactsContract + CalendarContractPlugin abstractions
Memory Web CanvasJetpack Compose Canvas + GesturesFlutter CustomPainter + GestureDetector
UI Rendering Styles4 styles (Glassmorphism, Neumorphism, M3, Minimal)4 styles (Glassmorphism, Neumorphism, M3, Minimal)
Color Palettes10 curated palettes (Dark + Light)10 curated palettes (Dark + Light)
AI Co-PilotNeoWispCoPilotEngine (built-in)copilot_chat_sheet.dart (Groq API)
08 / VISUAL IDENTITY SYSTEM

4 rendering styles. 10 color palettes.

Rendering Styles

Glassmorphism (Default)

Translucent frosted glass containers, hairline borders (0.5dp), dynamic backdrop blur, glowing elevation gradients.

Neumorphism

Soft extruded tactile surfaces with dual directional light and dark inner/outer drop shadows.

Minimalist

Crisp flat surfaces, strong typography hierarchy, monochrome cards, thin high-contrast borders.

Material 3

Standard Google Material You design system with filled tonal surface containers and rounded corners (24dp).

Color Palettes

Midnight Cyan
Emerald Mint
Sunset Amber
Cyber Violet
Obsidian Dark
Slate Monokai
Rose Quartz
Neon Pulse
09 / PRODUCT PERSONAS

Who NeoNote is built for.

💼

Sales & Account Executives

Quick post-meeting voice debriefs. Client names and follow-up commitments automatically logged into contact relationship timelines. No CRM typing after every call.

🚀

Founders & Consultants

Capturing ideas while driving or walking. Raw rants transformed into structured summaries and action lists. Never lose an idea to a forgotten note.

🧠

ADHD & Neurodivergent Thinkers

Zero-friction voice capture without the paralysis of blank text documents or complex folder hierarchies. Just hold and speak.

⚙️

Project Managers & Engineers

Quick standup update voice notes. Extracted checkable to-dos with due dates, auto-synced to device calendar. DPR updates on the go.

10 / TIER BREAKDOWN

Free tier. Pro tier.

Free Tier

  • 3 notes / day (max 60 sec per note)
  • On-device local engine (Vosk/Whisper)
  • Basic rule-based cleanup
  • No contact linking or timelines
  • Basic memory web view
  • Local storage only

Pro Tier

  • Unlimited notes and recording length
  • Groq Cloud Whisper-Large-V3 (highest precision)
  • Full AI summaries & smart formatting
  • Automatic contact matching & relationship feed
  • Full memory web with co-occurrence analytics
  • Encrypted multi-device sync & cloud backup
11 / V4 FLUTTER ARCHITECTURE

Cross-platform. One codebase. Three targets.

HomeScreen

Primary voice recording hub. Mic orb with animated pulse, real-time liquid waveform visualizer (5 parallel sine waves), live transcript preview.

FeedScreen

Searchable chronological notes feed. Category chip filter pills, inline audio player, note detail bottom sheet.

MemoryWebScreen

Interactive force-directed knowledge graph. Flutter CustomPainter with PERSON/PLACE/TAG node bubbles and Bezier edge curves.

TasksScreen

Action items and follow-ups. Filter tabs (All / Open / Completed), checkbox toggles, due date badges.

SettingsSheet

Configuration panel: Groq API key input, theme palette selector (10 palettes), UI rendering style switcher (4 modes).

CoPilotChatSheet

AI assistant bottom sheet powered by Groq LLM. Preset buttons: Analyze Top Threats, Memory Search, Relationship Insights.

12 / FAQs

Questions about NeoNote?

What is the relationship between NeoNote, neoWisp, and the product on this page?

How does the Preview-Then-Refine data flow work?

What makes the SSIM-based visual diff unique vs. standard note apps?

What is the difference between the Free and Pro tiers?

ARIVU PRODUCT - BENGALURU

Stop losing ideas to silence.

NeoNote is available for licensing or custom enterprise deployment. Contact Arivu Technologies to discuss integration into your organization's workflows.