Paid AI interview copilots charge a monthly subscription for something you can run yourself for free. This guide sets up AI Interview Helper, a free and open-source AI interview assistant, from zero to a working answer.
What you need
- A Windows 10/11 computer (on macOS, run it from source with Node.js 18+).
- About 5 minutes.
- A Google account to get a free Gemini key.
Step 1. Download and install
Download the Windows installer, or the portable .exe from the releases page if you don't want to install anything.
"Windows protected your PC"? The app isn't code-signed yet, because certificates cost money and the project is free. Click More info → Run anyway. The full source code is on GitHub if you want to check it.
Step 2. Get a free AI key
- Go to aistudio.google.com/apikey and sign in with your Google account.
- Click Create API key.
- Copy it. It starts with
AIza.
One Gemini key covers everything: chat answers, speech recognition and screenshots of coding tasks. The free tier is enough for practice.
Want even faster answers? Add a free Groq key as well. Groq is the fastest provider and uses Whisper for speech recognition, but its free models can't read screenshots, so keep Gemini for those. If Gemini isn't available in your country, Groq or Z.ai are good free options, and Ollama runs a model on your own computer without any key.
A note on privacy: on Google's free tier, your prompts may be used to improve Google's products. If that matters to you, use a paid key or a local model with Ollama.
Step 3. Paste the key
Open the app, click the settings icon (⚙) and paste the key into the Gemini field. A green dot means the key is set. Optionally, add the role you're interviewing for and paste your resume: answers then come in first person, based on your experience.
Step 4. Let the app hear the interviewer
Your microphone only hears you. To capture the other side of a Zoom, Google Meet or Teams call, add a system-audio source:
Then open the Audio tab, turn on "Distinguish interviewer's voice from mine" and set system audio as Interviewer and your microphone as My mic. The app will only answer the interviewer.
Step 5. Try it
| Shortcut | What it does |
|---|---|
Ctrl+Shift+L | Start or stop listening |
Ctrl+Shift+Space | Screenshot the screen and solve the task |
Ctrl+Shift+H | Hide or show the window |
Ctrl+Shift+C | Clear the chat |
Play a recorded interview question, or ask a friend to read one, and watch the answer appear when the speaker stops talking.
Which AI model to choose
| Goal | Good choice |
|---|---|
| Fastest answers | Groq · GPT-OSS 20B |
| Hard coding and system design | Claude Opus, or Groq · GPT-OSS 120B |
| Screenshots of tasks | Gemini Flash, Claude or GPT-4o (vision models) |
| Free and generous limits | Gemini Flash-Lite, Z.ai GLM-4.5-Flash |
| Fully offline | Ollama with a local model such as Llama or Qwen |
Add keys for two or three providers. If one hits a rate limit, the app automatically moves to the next one.
Troubleshooting
- No answer after speaking: check the volume meter in the Audio tab and lower the silence threshold.
- "All providers unavailable": check the API key, or add a second provider as a fallback.
- Screenshot answers ignore the image: pick a vision-capable model.
Ready to set it up?
It's free, open source and takes about five minutes.