Live annotations and screenshot markup

Needs

An active recording. No account.

Produces

Numbered annotations with comment, element description and screenshot

Shortcut

Enter saves, Shift+Enter adds a line, Esc backs out

When to use it

Recordings show what happened. Annotations say what was wrong about it. A QA tester can pin the misaligned total, the label with the wrong copy, or the region where a spinner never stops, without stopping to explain the surrounding steps. Developers then read the concern first and the diagnostics second.

Annotating the live page

1. Turn on Annotate

   While recording, toggle Annotate. Elements highlight in blue as you move the pointer, and normal page clicks pause.

2. Pick an element or an area

   Click a highlighted element, or drag a freeform rectangle across any region when the problem is not a single element.

3. Write the comment

   Type into the anchored comment bar and press Enter or the check button. A numbered pin marks the spot. If the recording microphone is on, speech fills the comment as you talk; the comment’s mic button starts or stops that microphone.

4. Get back to the page

   Choose Turn off annotations or toggle Annotate again. Esc first closes the comment, then leaves annotation mode. Pausing or stopping the recording also ends it.

Screenshot markup

Markup opens a separate drawing editor on a frozen screenshot of the tab. It has pen, arrow, rectangle, ellipse, text and highlight tools. Use it when you want to draw relationships, such as an arrow from a button to the panel it should have opened, rather than leave a comment on one thing.

What the report does with them

• Annotations appear with their pin number, comment, element description and screenshot, at their point on the timeline.
• When AI summaries are configured, annotations and notes are placed first in the evidence the model reads, so the issue title and summary describe your concern rather than unrelated console noise. See AI issue summaries.

Limits

• Live annotation is available only while recording. Screenshot markup remains available for captured frames.
• Spoken comments need a transcription provider. Deepgram and ElevenLabs fill the comment live; OpenAI fills it in segments, so text appears after a short pause. The audio stays part of the session’s voice recording.
• With an AI model configured, selecting an element sends a short excerpt of visible page labels to that model to build a vocabulary list for the transcription provider. See page vocabulary.