← Back to the interview lab

Setup & design notes

Aria needs no API key, no account and no installation. Everything below is optional polish.

1. Get started in 30 seconds

Pick a job family, type the organization and target role, paste the job description if you have one, choose HR / Technical / Behavioural / Case, then press start.

You can answer entirely by typing — voice and camera are optional.

2. Voice & camera permissions

Tap the mic to speak your answer; tap it again when you are done and it is sent automatically. Turn on Hands-free mode and the mic reopens after every question, like a real live call.

Speech input uses your browser's built-in recognition — best supported in Chrome, Edge and Safari. The first mic or camera tap raises a browser permission prompt; if you blocked it earlier, re-enable it from the padlock icon in the address bar.

Audio and video never leave your device: only the text transcript is sent for the next question.

3. How the AI stays secure

All model calls run through a server function on this app's backend, so the AI credential is never present in browser code, network requests or page source. There is nothing for you to paste or configure.

Questions are grounded with live web search when the model needs current facts about the company or industry, so references stay reasonably up to date.

4. Coaching signals

After every answer you get scores for clarity, structure, relevance and confidence, plus a one-line tip. They are directional coaching, not a hiring verdict. Press End & get debrief for a spoken wrap-up with strengths and improvements.

5. Design prompt

"Dark glass interview studio. Near-black teal background with soft emerald and cyan radial glows, frosted translucent cards with hairline emerald borders, one emerald accent for all actions, Sora display type, generous spacing, quiet motion — pulsing mic ring, breathing live dot, smooth meter fills. Calm and clinical, never gamified."

Tokens live in src/styles.css as oklch variables with glass, glow and text-gradient utilities.

6. Realtime voice roadmap

Aria today pairs a reasoning model with on-device speech, which works in every browser without keys. A fully bidirectional realtime audio session (streaming speech-to-speech over a WebSocket, authenticated with short-lived ephemeral tokens minted server-side so the permanent key stays on the server) is the natural next upgrade and would slot in behind the same UI.