Audio Response Survey Questions

Audio response questions let respondents answer by recording or uploading audio instead of typing. They are useful for open-ended interview-style prompts, oral history and language research, accessibility for respondents who find typing difficult, and studies where tone, hesitation, or emotion carries information a text box would lose.

Use audio response when hearing the answer matters. If a written transcript is enough, Text Response or Text Chat is easier for respondents to complete and for you to analyze at scale.

• In-browser recording: respondents record directly with their microphone, with a live level meter so they can see the mic is actually picking up their voice — no extra software needed.
• File upload: respondents can instead attach an existing audio file.
• Maximum duration: caps recording length so responses stay a manageable size.
• Instructions field: tells respondents what to talk about and how long to speak.
• Accessible text alternative: an optional, low-key "type instead" link for respondents who genuinely can't provide a spoken answer — self-service, no justification required, off by default so it isn't a general opt-out. The typed-answer box has its own Character limit, Disable pasting, and Track pasting settings. A required question accepts the typed answer in place of a recording.
• Optional AI transcription: an author can transcribe an individual response's audio from the Export tab, using their own speech-to-text API key.

1. Step 1: Open Block Builder. Open the survey and select the block where the spoken response belongs.

   Step 1: Open Block Builder

2. Step 2: Add an Audio Response question. Choose Audio Response from the question type selector.

   Step 2: Choose Audio Response question type

3. Step 3: Write clear instructions. Tell respondents what to talk about and roughly how long to speak, since there is no text prompt to re-read while recording. Instrument Review flags an Audio Response question left without instructions.

   Step 3: Write instructions

4. Step 4: Set the maximum duration. Choose a limit long enough for a real answer but short enough to keep file sizes reasonable; recording automatically stops at the limit.

   Step 4: Set maximum duration

5. Step 5: Choose recording, upload, or both. Allow upload as a fallback for respondents whose browser or device blocks microphone access. Instrument Review warns if both recording and upload are turned off, or if the minimum length is longer than the maximum.

   Step 5: Recording and upload options

6. Step 6: Consider the accessible text alternative. Turning on "Allow a typed-text alternative" shows a small, low-emphasis link a respondent can use to type their answer instead — meant for someone who genuinely can't provide audio, not a general opt-out, so it's off by default. A required Audio Response question with this left off is flagged by Instrument Review, since spoken-only items have no path forward for a respondent who can't record.

   Step 6: Recording and upload options

7. Step 7 (optional, beta): Set up AI transcription. Turn on "AI transcription" and connect a speech-to-text API key (or a saved credential) from a supported provider. This only makes a "Transcribe" action available per response in the Export tab — it never transcribes automatically, and nothing is sent anywhere until an editor clicks it for a specific response. The key and credential are never sent to respondents or included in the published survey.

8. Step 8: Preview and grant microphone access. Test both the record and stop controls, and confirm playback works before publishing.

   Step 8: Preview the recorder

Each response exports as a downloadable audio file for that question (or as plain typed text, for a response that used the accessible text alternative). Audio is stored inline with the response rather than in separate cloud storage, so keep the maximum duration reasonable — long recordings from many respondents increase the size of exported response data.

Response quality automatically flags two patterns specific to audio answers: a recording too quiet to plausibly contain an audible answer, and the exact same audio clip submitted for the same question by more than one response — the audio equivalent of a duplicate/canned open-text answer. A respondent who used the text alternative instead gets the same gibberish, repeated-text, and duplicate-answer checks any other open-text question already has.

A Silicon Sample run answers an Audio Response question as typed text when the typed-text alternative is on. Recording-only questions are left blank.

• File Upload Survey Questions for attaching an existing file instead of recording
• Text Chat Survey Questions for structured open-ended text
• Audio Conversation Survey Questions for a back-and-forth AI-moderated interview instead of a single clip
• Text Response Survey Questions
• Privacy and Anonymity in Research Surveys
• Prepare for IRB Research Survey Review