Voice notes, images and files in conversations

When a customer sends a voice note, a photo or a file in a conversation, STS Connect reads it for your team automatically. It works on every plan, with or without the AI agent, and reads up to three attachments per message.

Turn it on or off

In the portal open AI Agent and find the Media understanding section. It has three switches, all on by default:

• Transcribe voice messages: voice notes and the sound of videos become text.
• Read text in images and files: invoices, screenshots and PDF files your customers send are read.
• Describe images without text: photos with no printed words are described instead (see below).

What your team sees

The text appears as an internal note under the customer's message in the STS Connect Inbox. The customer never sees it. Each note says what it is, in your team's language: Voice message transcript · automatic, Video transcript · automatic, Text read from the image · automatic or Text read from the file · automatic. When a part may be inaccurate, the note says so.

When the AI agent handles the conversation, it then answers the text exactly like a typed message.

Images without text

A photo with no printed words cannot be read as text, but it can be described. Descriptions work only with your own AI key: the trial key never reads images, and the image is sent to your AI provider, never to ours. The description appears as an internal note labelled Automatic image description · may not be accurate, and the AI agent treats it as a hint, never as a fact.

On the same page, the Image model field chooses which of your provider's models sees images, and Test key checks it with a tiny test image:

• Your key accepts images: everything works.
• This model does not accept images; image conversations go to your team: pick a different image model.
• Your AI provider rejected the key: fix the key in the provider settings.
• The test could not finish. Try again.: a momentary problem; run the test once more.
• Enter an image model, or a chat model in the provider settings above, first.: nothing is sent until a model is set.

When the team steps in

If the model cannot see images, or media cannot be read at all, the conversation goes to your team with a note: The customer sent media that could not be read automatically; please review it. With Auto-reply to customers on, the customer is also told that a colleague takes over. With it off, the agent never writes to customers, so only your team sees the note.

Recent voice notes, images and files in the same conversation (within its last 15 messages) stay part of what the AI agent reads, so it can answer a follow-up question about them.

Monthly voice minutes

Voice notes and videos count against a monthly minutes allowance per plan; images and files never count minutes. The AI Agent page shows where you stand: This month: X of Y minutes. When the minutes run out, that message is not transcribed and your team gets a note saying so. The allowance resets every month.

	Free Trial	Basic	Professional	Enterprise
Voice minutes per month	60	300	1000	3000

See Plans and billing for everything else each plan includes.

Privacy

Reading runs on our own servers, which have no internet access, and no copy of the file is kept after it is read. The text is saved in the internal note your team sees. When the AI agent answers the conversation, the text goes to its AI provider like any typed message. An image itself leaves our servers only to be described, and only to the AI provider of your own key.

To feed the AI agent your own documents instead, see Upload files to the Knowledge Base.