Voice & Transcription
Choose audio inputs, use automatic or manual submission, and fix common transcription issues.
Voice & Transcription
Follow questions live and get answers based on your selected interview materials. Auto follows the conversation; Manual lets you choose when to submit.
Connect your audio
- Open voice controls. System Audio captures the other person in the meeting; Microphone captures you.
- Press
Cmd/Ctrl + Mto start and wait for the connection. - Ask the other person to say a sentence and check that text appears.
Use headphones to prevent meeting audio from reaching the microphone a second time. Keep the default Smart route unless you have a reason to change it.
Auto or Manual?
Use the Auto / Manual switch in the voice panel.
| Situation | Submission mode |
|---|---|
| Interviewer-led questions | Auto submits when a complete question is detected |
| Long questions or interruptions | Manual: press Cmd/Ctrl + Enter when ready |
| You ask a question yourself | Submit manually, or change the auto-submit scope below |
To trigger answers from your own questions, select Advanced Settings → Auto-submit scope → Interviewer + candidate. This applies only in Auto mode. The default scope is interviewer only.
Which question was submitted?
The highlight shows which question the answer belongs to. A question may span several transcript bubbles.
| Bubble | Meaning | What to do |
|---|---|---|
| “Transcribing” with changing text | Speech is still being captured | Wait for the complete question |
| Highlighted border or pulsing glow | Submitted question; answer is being generated | Check the answer panel |
| Highlight remains without pulsing | Answer completed | Continue with the next question |
| Orange warning border | This answer failed | Read the error in the answer panel, then retry |
Light mode uses blue emphasis; dark mode uses warm gold. Orange indicates failure. These markers show submission state, not accuracy or a score.
See the transcript and answer panels.
A new question arrives before the answer finishes
- Confirm the new question is in the transcript and is no longer temporary “Transcribing” text.
- In Voice Ask, press
Cmd/Ctrl + Enterto submit the new question. - The client interrupts the old answer request and starts the current question. Check that the highlight moves to the submitted transcript.
This also works in Auto mode. Do not use Cmd/Ctrl + G, which starts a new session. Wait for the question to finish before submitting, and avoid repeated presses; manual submission still checks that usable content is available.
Tailor your answers
- Select confirmed personal, job, and supporting materials under Current Interview. See Interview Materials.
- Set the response language, length, and additional instructions in Answer Preferences.
- Press
Cmd/Ctrl + Dto choose General, Coding, System Design, or Behavioral, then submit.
English-to-Chinese transcript translation
Enable Live translation under Advanced Settings → Voice Settings. The next time you start transcription, Chinese translations appear below both speakers' English speech. Available on the Flagship plan. Turning it off restores the previous recognition setting.
Troubleshooting
| Problem | Check first |
|---|---|
| You hear the meeting but see no transcript | Enable System Audio and its recording permission |
| Only the other person's speech appears | Enable Microphone input and permission |
| Sentences appear twice | Use headphones to prevent duplicate microphone capture |
| Your question is transcribed but not submitted | Submit manually or include candidate speech in the auto-submit scope |
| Answers start before the question ends | Switch to Manual; see Advanced Settings for silence timing |
| Connection fails | Check network, permissions, and available quota; follow the displayed error |
Start, Pause, Resume, and End
| Control | Effect |
|---|---|
| Start | Connect and begin transcription |
| Pause | Pause transcription until you resume |
| Resume | Continue the paused voice session |
| End | Finish the voice session |
Pausing transcription is different from Manual submission. Manual keeps listening and transcribing, but waits for you to request an answer.
Speech recognition, custom vocabulary, and pauses
Start with Smart routing. Choose a specific speech service only if Smart routing is not working well on your network. Add hotwords for misrecognized company names, project names, or acronyms. Silence timing affects transcript splits; auto-submit scope controls which roles can trigger answers.
Switch views or finish
- For on-screen code or questions, press
Cmd/Ctrl + Hto capture anytime with Screenshot Ask. PressCmd/Ctrl + 1to return to Voice Ask. - Screenshots do not carry into Voice Ask automatically. Keep questions that depend on the image follow-ups in Screenshot Ask.
- Use voice controls to pause or end the session.
Cmd/Ctrl + Gstarts a clean session;Cmd/Ctrl + Ronly resets the window layout.
Something missing or unclear?
Tell us which page or step needs work.
AI Reading Assistant
Send to your preferred AI
Smart Summary
Deep Analysis
Key Topics
Insights