Voice, files and screenshots

Customers can talk to the AI support agent out loud, in the same widget they chat in.

• Talk uses the same Knowledge, signed-in customer context and handoff as chat.
• Speech is detected automatically and Voice answers out loud in 40+ languages; where spoken output isn't supported, the conversation continues in text.
• Voice minutes are included in paid plans. The free trial does not include Voice minutes.

1. 1Open Channels -> Widget and turn on Talk (Enable Talk widget).
2. 2Choose a voice from the voice catalog and preview it.
3. 3Write the welcome message the voice plays when a customer starts Talk, and pick the orb texture.
4. 4Optionally make Talk the surface the widget opens on: Home, Chat or Talk.

• Voice notes customers send on messaging channels are transcribed.
• The agent answers from the transcript, and your team can read it in the thread.
• The original audio stays in the conversation and can be played back.

• In the widget, customers can attach images, screenshots and photos, and PDF, DOC, DOCX and TXT files, up to 8 MB per file.
• Images and screenshots are read with vision, so the agent can answer from what is on screen, such as an error message.
• PDF and Word documents are parsed, including tables.
• Photos and documents sent on WhatsApp, Instagram and Telegram are read as context in the same way.
• You can turn widget uploads off in Widget settings.

• When a customer pastes a link, the agent fetches the page safely and reads it as context for that conversation.
• Customer links are context, not your approved knowledge. Answers about your product still come from your Knowledge.

• Images can be previewed directly in chat.
• Documents can be opened or downloaded depending on format.
• Voice notes include playback controls and related message context.