Prerequisites

Intervu requires external dependencies to capture audio, transcribe speech, and generate answers. Setup differs between Windows and macOS.

Overview

Component	Purpose	Windows	macOS
Virtual Audio	Capture system audio	VB-Cable	BlackHole
FFmpeg	Audio processing	Bundled	Manual install (brew install ffmpeg)
Speaches	Speech-to-text endpoint	✅ Required	✅ Required
LLM	Answer generation	✅ Required	✅ Required

---

1. Virtual Audio Device

Intervu captures system audio (interviewer's voice from video calls) using a virtual audio device.

Windows: VB-Cable

VB-Cable acts as a virtual audio cable:

• Input (CABLE Output): Route your system audio here
• Output (CABLE Input): Intervu captures from here

Installation

1. Download VB-Cable from vb-audio.com/Cable
2. Extract the ZIP file
3. Run VBCable_Setup_x64.exe as Administrator
4. Reboot your computer
5. After reboot, you'll see "CABLE Input" and "CABLE Output" in your audio devices

Configuration

1. During your interview, set your video call's speaker output to CABLE Input
2. In Intervu Settings, select CABLE Output as your System Audio Device
3. Your microphone goes directly to Intervu as the Microphone Device

You Won't Hear System Audio Through VB-Cable Alone

When you route system audio to VB-Cable, it captures the audio for Intervu — but you can't hear it through your speakers. To hear the interviewer while Intervu captures the audio, use one of these audio routing solutions:

• SteelSeries GG (steelseries.com/gg) — Free audio routing software that lets you mirror audio to multiple outputs. Set it up to send audio to both your speakers and VB-Cable.
• Voicemeeter (vb-audio.com/Voicemeeter) — Virtual audio mixer that can route audio to VB-Cable and your speakers simultaneously. More powerful but has a steeper learning curve.

With either solution, the interviewer's audio goes to your speakers (so you can hear them) and to VB-Cable (so Intervu can capture it).

Simple Flow (VB-Cable Only — You Can't Hear Audio)

┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│ Video Call  │────▶│ CABLE Input  │────▶│ CABLE Output│
│ (Speaker)   │     │ (VB-Cable)   │     │ (Intervu)   │
└─────────────┘     └──────────────┘     └─────────────┘
                          │
                          ✗ You can't hear the audio!

Recommended: VB-Cable + SteelSeries GG or Voicemeeter (You Can Hear Audio)

Using SteelSeries GG:

1. Install SteelSeries GG
2. In SteelSeries GG, set up audio mirroring to send audio to both your speakers and VB-Cable
3. Your video call audio now goes to speakers (you hear it) and VB-Cable (Intervu captures it)

Using Voicemeeter:

1. Install Voicemeeter
2. Set Voicemeeter as your default audio output
3. In Voicemeeter, route audio to both your speakers (A1) and VB-Cable (A2)
4. Intervu captures from VB-Cable as before

┌─────────────┐     ┌──────────────────┐     ┌──────────────┐     ┌─────────────┐
│ Video Call   │────▶│ SteelSeries GG / │────▶│ CABLE Input  │────▶│ Intervu     │
│ (Speaker)   │     │ Voicemeeter      │     │ (VB-Cable)   │     │ (Capture)   │
└─────────────┘     └──────┬───────────┘     └──────────────┘     └─────────────┘
                           │
                           ▼
                    ┌──────────────┐
                    │ Speakers /   │
                    │ Headphones  │
                    │ (You Hear)  │
                    └──────────────┘

macOS: BlackHole Audio

On macOS, use BlackHole instead of VB-Cable. BlackHole is a free, open-source virtual audio driver.

Installation

1. Download BlackHole from existential.audio
2. Open the .pkg file and follow the installation prompts
3. Reboot your Mac

Create a Multi-Output Device

Unlike VB-Cable on Windows, BlackHole alone routes audio away from your speakers. You need a multi-output device to hear audio while Intervu captures it:

1. Open Audio MIDI Setup (Cmd + Space → "Audio MIDI Setup")
2. Click + → Create Multi-Output Device
3. Check BlackHole 2ch and your speakers/headphones
4. Rename it (e.g., "Intervu Multi-Output")
5. Open System Settings → Sound → Output
6. Select "Intervu Multi-Output"
7. In Intervu Settings, select BlackHole 2ch as System Audio Device

For detailed macOS setup instructions, see macOS Setup Guide.

Why BlackHole + Multi-Output?

The multi-output device sends audio to both your speakers (so you can hear the interviewer) and BlackHole (so Intervu can capture it). This solves the same problem that SteelSeries GG/Voicemeeter solve on Windows — hearing system audio while capturing it.

---

2. FFmpeg

FFmpeg handles audio capture and device enumeration.

Windows

FFmpeg is bundled with Intervu on Windows. It's automatically downloaded during installation. No manual setup required.

macOS

FFmpeg must be installed manually on macOS:

brew install ffmpeg

See macOS Setup for more details.

macOS Requires Manual FFmpeg

Unlike Windows, FFmpeg is not bundled with the macOS version. You must install it via Homebrew before using Intervu.

---

3. Speaches (Speech-to-Text)

Intervu requires an OpenAI-compatible STT endpoint. We recommend Speaches — a self-hosted Whisper transcription service.

Quick Start

# Download Docker Compose files
curl --silent --remote-name https://raw.githubusercontent.com/speaches-ai/speaches/master/compose.yaml
curl --silent --remote-name https://raw.githubusercontent.com/speaches-ai/speaches/master/compose.cpu.yaml

# Set compose file
export COMPOSE_FILE=compose.cpu.yaml

# Start Speaches
docker compose up --detach

# Download a transcription model
export SPEACHES_BASE_URL="http://localhost:8000"
uvx speaches-cli model download Systran/faster-distil-whisper-small.en

For detailed instructions, see Speaches Setup.

Default Endpoint

• URL: http://localhost:8000/v1/audio/transcriptions
• Model: Systran/faster-distil-whisper-small.en

---

4. LLM Endpoint

Intervu generates suggested answers using any OpenAI-compatible LLM.

Options

Option	Type	Endpoint	Notes
Ollama	Local	http://localhost:11434/v1/chat/completions	Free, runs locally
OpenWebUI	Local/Remote	https://your-instance/api/chat/completions	Your private instance
OpenAI	Cloud	https://api.openai.com/v1/chat/completions	Requires API key, costs money

Recommended: Ollama

# Install Ollama from ollama.ai
ollama pull llama3.2:3b
ollama serve

For detailed instructions, see LLM Setup.

---

Verification Checklist

Before using Intervu, verify:

Windows

• [ ] VB-Cable installed and visible in Sound Settings
• [ ] (Recommended) SteelSeries GG or Voicemeeter configured for audio routing
• [ ] Speaches running (docker ps shows the container)
• [ ] STT model downloaded (Systran/faster-distil-whisper-small.en)
• [ ] LLM endpoint running (Ollama, OpenWebUI, or have OpenAI API key)

macOS

• [ ] FFmpeg installed (ffmpeg -version works)
• [ ] BlackHole installed and visible in Sound Settings
• [ ] Multi-output device created in Audio MIDI Setup
• [ ] Multi-output device set as system sound output
• [ ] Quarantine removed (xattr -d com.apple.quarantine /Applications/Intervu.app)
• [ ] Microphone permission granted
• [ ] Speaches running (docker ps shows the container)
• [ ] STT model downloaded
• [ ] LLM endpoint running

Next Steps

• Configure Audio Devices
• Configure STT Endpoint
• Configure LLM Endpoint
• macOS Setup Guide — Detailed macOS instructions