KnowGhost

Your smart free assistant for interviews and exams

✦ Key Features

🎤 Speech & Transcription

• System audio capture — captures audio directly from your system (Zoom, Teams, browser, any app) without microphone feedback
• Microphone support — also works as a classic voice recorder
• Multi-provider transcription — OpenAI Whisper, Google AI, Deepgram, or local whisper.cpp (CPU or GPU/CUDA)
• One-click whisper.cpp installer — downloads and sets up whisper.cpp automatically, no manual configuration
• Customizable transcription prompt — tuned for technical interviews, preserves English terms (APIs, Redis, Postgres, etc.)

🤖 AI & Chat

• LLM-agnostic — works with any OpenAI-compatible endpoint: DeepSeek, OpenRouter, Google AI, local Ollama, TabbyAPI, or your own custom provider
• Multimodal chat — paste images directly into messages (Ctrl+V), the LLM sees and understands them (diagrams, code screenshots, whiteboards)
• Screenshot analysis — capture screen region or full display and send to a dedicated vision model for analysis
• Adjustable system prompt — customize the AI's behavior, or use the built-in prompts that adapt to your UI language
• Conversation history — full chat history preserved, scrollable and searchable

💾 Card Storage

• Save any conversation — press one button to save a full Q&A session into a searchable card
• Auto-save — when chatting inside a card, messages are saved automatically — no manual saving needed
• Search & filter — find past conversations by title or tags
• Summary cards — one-click summarization of any conversation into a concise card

👁 Stealth & UI

• Invisible to screen capture — the window is excluded from screen recording and screen sharing (Zoom, Google Meet, Teams, OBS, etc.)
• Always-on-top — floats above other windows, with adjustable transparency for minimal intrusion
• Dark & light themes — MUI-based with orange accent (#db704b)
• i18n — English and Russian interface

⚡ Other

• Hotkeys — customizable shortcuts for quick audio capture and stream sending
• Local AI — run both LLM and transcription fully offline for privacy and zero latency
• Cross-platform — Windows, macOS, Linux

▸ How to Use

1. Setup

1. Open KnowGhost
2. Go to Settings → AI
3. Enter your API key (OpenAI, Google AI, or any OpenAI-compatible provider)
4. Choose LLM Model
5. Choose audio input:
   • System Audio — capture sound from apps (Zoom, Teams, etc.)
   • Microphone — record your voice
6. Configure Transcription model and prompt

2. Usage

1. Switch to Main tab
2. Click Start Audio Loop — recording in the background
3. When needed, press hotkey or Send Last X Seconds to get an AI response
4. Get instant answers during interviews or exams

3. Tips

• Use system audio to capture interviewer's questions
• Adjust opacity and always-on-top for stealth
• Configure hotkeys for quick actions
• Practice before important events

⊕ Local Setup

KnowGhost supports fully local operation — no cloud dependency, maximum privacy.

Local Speech Recognition

1. In settings: Mode → Transcription = Local
2. Choose Source = Managed (one-click install) or External (your own whisper-server)
3. For managed: pick variant (CPU or GPU/CUDA), click Install
4. The app downloads and sets up whisper.cpp automatically

⬇ Download

Download the latest release

---

Made with ❤️ for successful interviews and exams