Skip to main content

How evaluations work

This article explains how SpeechGradebook evaluates a student speech. It applies to instructors who want context before running evaluations.

Evaluation flow​

  1. You upload a recording and select a course, student, and rubric.
  2. The app sends the video and rubric to the SpeechGradebook Text + Video Model (Qwen) through the backend at /qwen-api/*.
  3. The model analyzes the recording against your rubric criteria.
  4. Results return scores, category feedback, transcript text, and timeline-style observations.
  5. You review, edit, and save the final record.

Processing time depends on video length and current service load.

Instructor review​

AI-generated scores and comments are a starting point. You can edit letter grades, points, comments, and timeline markers before saving. Your edits are stored as the official evaluation record.

Results may be held locally until the student completes consent. See Consent and data storage.

Related: Run a single evaluation