Use images, files, and voice

Add reference material, explain a screenshot, diagnose a problem, or dictate your next request.

Attach a useful image

Attach a reference image or screenshot and explain what to notice. Supported formats are PNG, JPG, WEBP, and GIF, up to three images. Each source must be no larger than 12 MB. Images are resized and the prepared set must fit within 6 MB.

Project chat offers Explain for interpretation, Diagnose for investigation, and Fix for a code change. Choose the intended action and review before sending. Analysis and edits can incur usage charges.

Attach a short reference file

You can attach up to three supported text or code files, with a maximum of 10 KB each and 20 KB combined. Supported extensions include TXT, Markdown, JSON, JavaScript, TypeScript, CSV, YAML, HTML, CSS, and XML. Empty, binary, and detected credential content is rejected. PDF documents and large archives are not supported by this attachment control.

Attach a short FAQ, sample data, design notes, or code example and explain its purpose. A builder attachment does not automatically become searchable knowledge for customers.

Dictate, review, then send

  1. In the project composer, select the microphone and allow browser microphone access.
  2. Record your request, then stop. Recording ends automatically at 90 seconds.
  3. Wait for transcription, correct names and instructions in the text, then send it yourself.

Transcription is a separate operation and may be billed even before you send the resulting request. If microphone permission or recording support is unavailable, type the request instead. Builder dictation does not add voice-note support to your customer bot; request that ability separately.