Verbal Victory
Real-time AI coaching for how you speak
An AI communication coaching platform. Record a practice conversation, get a scored breakdown of clarity, confidence, pace, and filler words, then drill the weak spots. The roadmap centers on a screen-capture-proof desktop overlay that coaches you during real meetings without your audience ever seeing it.
The Challenge
Communication coaching is expensive, slow, and absent at the moment it matters most: while you are actually speaking. Existing AI tools analyze you after the meeting ends, and their coaching interfaces show up in screen shares.
Human coaching does not scale and cannot attend your meetings
Competing tools like Yoodli deliver feedback only after the conversation is over
Teleprompter and coaching overlays appear in screen shares, making them unusable in real presentations
Feedback that arrives days later rarely changes in-the-moment habits like pace and filler words
People avoid practicing hard conversations because there is no safe, private place to rehearse
The Solution
A three-pillar platform built solo on a modern web stack. Pillar one ships today: record a scenario-based practice session and get an instant, scored analysis. Pillars two and three, a live in-session coach and an invisible meeting overlay, are designed and feasibility-tested.
Intelligent scoring: Whisper transcribes the session, GPT scores clarity and confidence, and local math computes pace and filler percentage into a weighted 0-100 score
Actionable feedback: every session returns three strengths, three improvements, and three practice tips specific to the scenario
Scenario library: job interview, public speaking, conflict resolution, sales call, and life partner conversations with difficulty progression
Tiered model selection: GPT-5 family models matched to user tier, with cost tracking, response caching, and usage limits built into Supabase
Real-time coach (designed): live transcript with filler highlighting, visual prompts like slow down or filler alert, and low-latency voice nudges via Cartesia or ElevenLabs
Invisible overlay (researched): Tauri desktop app using Windows screen-capture exclusion so coaching prompts stay visible to you and invisible to Zoom, Teams, and OBS
The Results
Phase 1 is complete and working end to end: recording, transcription, scoring, session history, auth, and tier enforcement. The honest picture as of June 2026 is a production-ready build awaiting launch, with the differentiating overlay validated on paper and scoped as a Windows proof of concept.
Working Scoring Pipeline
Audio in, weighted 0-100 score out: Whisper transcription, GPT analysis, instant pace and filler math, all persisted to session history
Unit Economics Documented
Roughly $0.15-0.50 in AI cost per analyzed session, with caching and tier-based model selection keeping the floor low
Overlay Feasibility Proven on Paper
January 2026 research confirmed Windows screen-capture exclusion works with Zoom, Teams, and OBS, and honestly documented that macOS 15+ blocks it
Validated Market Thesis
Yoodli reached a $300M+ valuation and Deepgram acquired Poised for real-time coaching tech, confirming the category Verbal Victory targets
Gallery
Technical Architecture
React and Vite frontend with Supabase for auth, Postgres with row-level security, and storage. Netlify Functions run the analysis pipeline: Whisper transcription, local pace and filler calculation, then GPT scoring with results written back to session history. The designed real-time path adds the OpenAI Realtime API over WebSocket, and the overlay path adds a Tauri desktop app with content protection enabled.
Technology Stack
