Media and technical question types

Audio and video recording

Audio recording captures a spoken answer, which suits language checks, call-handling scenarios, and any prompt where verbal delivery matters. Video recording adds the camera, so use it when on-camera presence is genuinely part of the skill you are assessing — a pitch, a briefing, a customer-facing explanation.

In the item editor you set a recording time limit and how many recording attempts the candidate gets, so a stumbled first take does not have to be their only take. Microphone and camera access for recorded answers is requested in the browser when the candidate starts recording; when the delivery also uses camera proctoring, screen recording, or ID verification, a consent and preflight gate verifies the setup before the exam begins.

Photo capture and file upload

Photo capture asks the candidate to take a picture, which is useful when the evidence is physical or visual. File upload accepts a produced artifact — a document, a spreadsheet, a design file — making it the natural type for work-sample tasks where you review the actual deliverable. The item editor lets you restrict the allowed file types and set the maximum number and size of files.

If the delivery uses fullscreen lockdown, note that an item's own file picker has a grace period: choosing a file through the question's picker does not count as a violation, so candidates can upload their files without triggering integrity events.

Code questions

The code type is built for programming tasks. Use it when you want candidates to write code rather than describe it, so evaluators can review the actual implementation as evidence. Combine it with a sidebar layout to keep a specification or bug report visible next to the answer area while the candidate works.

What review looks like

None of these types carries an answer key, so they auto-score as null — meaning "needs manual review", never zero. Evaluators review the recording, image, file, or code and score it, typically against a shared rubric, with notes attached as evidence and every score attributed to a named evaluator with a timestamp.

For long responses, AI assistance can summarize the material and draft criterion-level score suggestions, but every AI output carries model and version provenance, and a named person confirms or overrides it. AI supports the review; it never makes the decision.