v1.0.0 Production Release

Universal Stealth Interview Copilot & Speech Intelligence

Sub-second speech-to-text, adaptive 9-domain reasoning frameworks, and zero-distraction desktop overlay technology. 100% local, private, and Bring Your Own Key (BYOK).

STT Latency
~250ms
Groq Whisper Turbo
AI Providers
9 Engines
Direct HTTPS BYOK
Telemetry
0ms / None
100% Local Encrypted
Overlay Modes
3 Profiles
Studio • HUD • Notch

Engineered for performance & privacy

WishPilot was architected from the ground up to solve the three fatal flaws of existing interview tools: crippling latency, centralized privacy leaks, and generic responses.

Local-First & Zero Telemetry

No WishPilot central servers. Your audio buffers, candidate profile, resumes, and interview transcripts reside strictly in encrypted local storage.

Principle 01

Bring Your Own Key (BYOK)

Connect directly to official provider APIs (Groq, Cerebras, Together, NVIDIA, OpenAI, etc.) over direct TLS HTTPS. No middlemen or markup.

Principle 02

Sub-Second Latency Architecture

Custom Web Audio DSP pipeline buffers 16kHz audio and transcribes via Groq Whisper Large v3 Turbo in 180ms–350ms, allowing natural conversational scanning.

Principle 03

Adaptive 9-Domain Engine

Interviews differ across industries. WishPilot dynamically switches reasoning frameworks between Software, Finance, BPO, Sales, HR, PM, and Clinical roles.

Principle 04

High-performance desktop capabilities

Engineered exclusively for Windows 10 & 11 with native low-level display protection and multi-provider streaming.

Instant Answer Refinement

Immediately transform answers with 4 contextual pills: Make Shorter (15s pitch), More Technical (algorithms & metrics), Give an Example (real case studies), and Simpler Language.

4 contextual actions

Display Protection (WDA)

Enforces native Windows Display Affinity (WDA_EXCLUDEFROMCAPTURE). The overlay remains visible to you on your monitor, but completely invisible in Zoom, Teams, and browser screen shares.

Native Win32 API

Web Audio DSP Pipeline

High-efficiency AudioWorklet captures 16kHz audio, dynamically detects speech energy, filters silence, and streams audio directly to Groq Whisper Turbo for instant transcription.

180ms–350ms STT

9 unified AI providers

Direct HTTPS streaming with Bring Your Own Key (BYOK)

Zero markup, no proxy
Groq
Groq
Llama 3.3 70B & Whisper Turbo
~220ms
Cerebras
Cerebras
Llama 3.3 70B (Wafer-Scale Engine)
~1,800 tps
Together AI
Together AI
Llama 3.3 70B Turbo & DeepSeek
~450ms
Fireworks AI
Fireworks AI
FireFunction v2 & DeepSeek R1
~380ms
NVIDIA NIM
NVIDIA NIM
Llama 3.1 70B & Nemotron 70B
~410ms
Hugging Face
Hugging Face
Serverless Inference APIs
~600ms
OpenRouter
OpenRouter
Unified Gateway & Multi-Model
~550ms
OpenAI
OpenAI
GPT-4o & GPT-4o-mini
~650ms
Google Gemini
Google Gemini
Gemini 2.0 Flash
~390ms

Global keyboard shortcuts

Toggle Floating Stealth HUD / StudioCtrl + Shift + H
Mute / Unmute Live Speech EngineCtrl + Shift + M
Force Trigger Instant AI AnswerCtrl + Shift + A
Capture Active Screen / LeetCode ContextCtrl + Shift + S
Toggle Click-Through Transparent ModeCtrl + Shift + C
Quick Dismiss (Panic Hide & Mute)Ctrl + Shift + X

Multi-industry category engine

Interviews are never one-size-fits-all. WishPilot dynamically transforms its prompt reasoning, terminology, and evaluation frameworks across 9 distinct professional streams.

Active stream

IT & Software

Enforced framework
Distributed Systems & LeetCode Big-O
Framework focus points
  • Explicit Time & Space Complexity (Big-O analysis)
  • Distributed cache coherence, CAP theorem, and database isolation levels
  • System fault tolerance, circuit breakers, and async message queues
  • Production-grade error handling and edge cases
Expected tone of delivery

Architectural, concise, trade-off focused

Sample generated TL;DR
We use write-through Redis caching with Kafka-driven async invalidation to maintain sub-5ms reads with strict eventual consistency.
Configured in interviewCategories.jsxSeamless hot-switching supported

Technical architecture & pipeline

Engineered with a low-level desktop footprint, eliminating intermediaries to deliver the world’s fastest interview speech intelligence loop.

01

Audio Capture & DSP Buffering

Web Audio API • AudioWorkletNode

Microphone or system audio is captured at native rates and downsampled to single-channel 16kHz PCM in an isolated worker thread without blocking the UI.

< 20ms processing
02

Speech Intelligence Engine

Groq Whisper Large v3 Turbo

Client-side Voice Activity Detection (VAD) detects speech boundaries, dispatches audio buffers over direct HTTPS, and extracts technical transcripts.

180ms – 350ms STT
03

Context Synthesis & Prompt Adapter

Resume Parsing • Vision Context • Category Rules

Transcribed questions are merged with active candidate resume details, screen capture context (e.g. LeetCode problems), and active category frameworks.

Zero Telemetry
04

Unified Multi-Provider Inference

BYOK • SSE Streaming • 9 Providers

Streamed word-by-word via Server-Sent Events (SSE) directly from Cerebras, Groq, NVIDIA, Fireworks, Together, or OpenAI to eliminate latency bottlenecks.

Up to 1,800 tps
05

Stealth Display & Win32 Protection

SetWindowDisplayAffinity • WDA_EXCLUDEFROMCAPTURE

Native Windows API prevents the window from being rendered into graphics capture buffers, keeping your overlay 100% invisible on Zoom, Teams, and browsers.

Hardware Layer

Common questions & guidance

Everything you need to know about architecture, privacy guarantees, hardware compatibility, and deployment.

Yes. WishPilot is 100% free and open-source under the GNU General Public License v3.0 (GPL-3.0). There are no subscription paywalls, no monthly fees, and no feature limits. You bring your own API keys directly from official AI providers (many of which, like Groq and Cerebras, offer generous free tiers).

Educational & practice simulation disclaimer

WishPilot is engineered as a high-performance interview preparation simulator and real-time speech intelligence training tool. It is designed for self-directed mock interviews, technical communication drills, and live presentation pacing.

The author and contributors assume no liability for any unauthorized or non-compliant use of this software. Users are solely responsible for ensuring their usage strictly complies with all applicable institutional policies, professional ethics, corporate terms of service, and testing guidelines.