Free Text Input
Participants write descriptions, transcriptions, or reviews for each item. In unmoderated mode it's a simple textarea. In moderated mode, an AI voice guide reads what the participant wrote and asks adaptive follow-up questions — no two sessions are the same.
Candor does this so you don't have to
Independent writing
Each participant writes their response independently. Items are presented in randomized order. No one sees what others have written — eliminating groupthink and anchoring effects.
AI reads and probes
In moderated mode, the AI voice guide reads the participant's written response and asks adaptive follow-up questions. It probes deeper on interesting comments, explores edge cases, and validates intent in real time.
Transcription-aware timing
For audio and video items, Candor estimates 4x the media duration for replay and typing. Reward calculations ensure participants are fairly compensated for the actual time spent on each task.
Grouped by item
Responses from all participants are automatically grouped by item. You see every response for each asset side-by-side, making it easy to spot patterns and outliers across your dataset.
{ item: "img_01", responses: [3] }Audio Transcription
High-fidelity manual transcription with duration-based rewards calculated at 4x media length for fair compensation.
Image Description
Generate detailed alt-text and semantic descriptions for visual datasets at scale. Perfect for accessibility and training data.
Product Reviews
Collect qualitative feedback on early-stage prototypes or existing products with optional AI-guided follow-ups.
Adversarial Prompting
Testing AI safety by having human evaluators attempt to bypass constraints and document their strategies.
Document Annotation
Expert review of legal or technical documentation with clause-by-clause notes and structured commentary.
Open-ended Surveys
Move beyond multiple choice with deep-dive written responses. Moderated mode probes for richer, more actionable insights.