Voice Session and Transcription

Voice Session provides real-time transcription and AI-assisted answers for video and phone interview scenarios.

Start a Voice Session

Open Voice Control from the toolbar
Select Microphone or System Audio as input, then choose Smart route or a fixed voice route if needed
Click Start to enter live transcription

Before a real interview, run one test call to verify permissions and input source.

If you are not sure which route to pin, start with Smart route. Once you know which route performs best on your network and interview platform, you can lock it manually.

Session Controls

Action	Description
Start	Begin transcription and activate session
Pause	Pause current transcription
Resume	Continue from paused state
End	End and close current session

Transcription Status

Status indicators include:

Connecting: building speech service connection
Connected: transcription is running
Paused: session is paused
Error: network or quota issue

Long silence may trigger split or pause indicators depending on your settings.

Smart route, silence split, and hotwords

Feature	Best use case
Smart route	You want the system to choose a more reliable route automatically
Fixed route	You have already tested and know which route works best on the current platform
Hotwords	Company names, project names, acronyms, and technical terms are often misrecognized
Silence split	Shorter pauses call for lower thresholds; longer interviewer questions often benefit from higher thresholds

Practical tuning guidance:

If the round contains many acronyms or mixed-language technical terms, add hotwords first before changing other voice settings.
If the interviewer pauses mid-sentence a lot, avoid setting the silence threshold too low.
After every adjustment, rehearse once on the target platform or /tools.

Voice Q&A strategy

Message filter: interviewer-only / candidate-only / all
Recent message count: choose how many turns are analyzed (1-10)
Carry previous QA: reference prior turn in next answer
Use knowledge base: include material context in voice response, especially for project follow-ups, behavioral rounds, and introductions

Auto-triggered Q&A vs manual timing

Mode	Better for
Auto-triggered Q&A	Interviewer-led rounds where you want fewer actions. With `Interviewer only`, the system is better positioned to trigger after a complete interviewer question
Manual timing	Long questions, multi-speaker discussions, frequent interruptions, or any round where you want to wait until the question fully lands

Usage notes:

If the interviewer is doing most of the talking, try Interviewer only first. It usually creates a cleaner signal.
If the round is discussion-heavy or the question often arrives in multiple parts, manual timing is usually safer.
The knowledge base panel can stay open as a stealth-friendly note layer without blocking live voice Q&A.

Working with screenshot Q&A

Cmd/Ctrl + 1 switches to the live transcription Q&A view.
Cmd/Ctrl + 2 switches to the screenshot Q&A view when the interviewer shares code, diagrams, or a whiteboard.
If your knowledge base note panel is open, it can stay there as a reference layer alongside both voice and screenshot Q&A.

Recommended Practices

Confirm microphone permission before interview starts
If latency rises, first try Smart route or switch to the route you already validated
If network is unstable, switch to screenshot mode as fallback
Rehearse Smart route, hotwords, silence threshold, and auto-trigger rhythm before the live interview