Speak Savvy
Voice practice app with real-time feedback on clarity, pacing, and structure for interviews and presentations.
The Problem
Speaking skills degrade without practice, and practicing alone provides no feedback. You can rehearse an interview answer in the mirror, but you won’t catch filler words, pacing issues, or structural problems. Professional coaching costs hundreds per session. Recording yourself helps but requires manual review afterward.
What I’m Building
Speak Savvy is a mobile voice practice app that listens to your spoken responses and provides immediate feedback on clarity, pacing, filler word usage, and answer structure. You select a practice mode (behavioral interview, technical explanation, elevator pitch, presentation), receive a prompt, speak your response, and get scored feedback with specific improvement suggestions.
Practice Modes
Behavioral Interview — STAR-format prompts (Situation, Task, Action, Result). The app detects whether your response follows the STAR structure and flags missing sections. Common prompts like “Tell me about a time you handled conflict” with follow-up questions that probe deeper.
Technical Explanation — Explain technical concepts to different audiences. The app scores jargon density and suggests simpler alternatives when you’re explaining to a non-technical audience. Prompts range from “Explain how a database index works” to “Describe your system architecture”.
Elevator Pitch — Timed 30/60/90-second pitches with pacing feedback. The app tracks whether you’re front-loading key information or burying the lead. Scores hook strength, value proposition clarity, and call-to-action presence.
Presentation — Longer-form practice with section timing. Upload slide outlines and the app tracks pacing per section, flagging where you rush or drag. Transition detection scores how smoothly you move between topics.
Feedback System
Feedback combines speech recognition analysis with structural evaluation. Real-time indicators show speaking pace (words per minute), filler word detection (“um”, “like”, “you know”), and pause patterns. Post-response analysis scores clarity (vocabulary level vs audience), structure (logical flow, section completeness), and confidence signals (vocal variety, assertion vs hedging language).
Feedback cards summarize each dimension with a score and specific examples. “Clarity 8/10 — clear vocabulary, one jargon term (‘microservices’) could be simplified for this audience.” “Pace: Good — 142 WPM average, slight rush in final 15 seconds.”
Technical Approach
Built natively with Swift and SwiftUI for iOS. Apple’s Speech framework handles real-time transcription with word-level timestamps. AVFoundation manages audio capture with noise cancellation. NLP analysis runs on-device using Apple’s Natural Language framework for filler word detection and sentence structure analysis. More complex structural scoring uses a lightweight model for STAR detection and argument flow analysis. All audio processing happens locally — recordings never leave the device.