Some documentation gets read at a desk. Plenty of it gets consumed while someone is doing something else — commuting, cooking, walking between meetings, or halfway through a deployment with both hands busy.
FinalDoc knowledge bases can be listened to, and talked to. Those are two separate features, and it is worth being clear about which is which.
Listen to any article
Readers get a Listen button on articles. The narration is generated from the article itself rather than from a summary, so what they hear is what you wrote — headings, steps and all. Private notes are stripped before anything is spoken.
The player gives readers the controls you would expect: play and pause, seek, playback speed and volume. Narration uses a single clear narrator voice, chosen to suit reference material rather than to sound like a personality.
You control whether the button appears at all, with the Enable read out loud toggle in your article settings. Switching it off removes the feature — the button disappears and the narration endpoint stops serving, so it is a real off switch rather than a hidden one.
Ask questions out loud
The more interesting half is conversational. A reader opens the assistant in voice mode and simply talks. The assistant answers out loud, in real time, and can be interrupted mid-sentence the way a person can — start speaking and it stops to listen.
Both sides of the conversation are transcribed into the panel as it happens, so a reader can read back what was said and copy anything useful.
Readers pick the voice themselves from ten options in the panel, each with an instant preview before they commit. Switching voice mid-conversation reconnects without dropping them out of voice mode.
Grounded in your documentation
The answer is grounded in your knowledge base, not in whatever the model happens to know about your product category. When a reader asks something, the assistant searches your published articles and reads from the ones that match — pulling real article text, not a cached summary of it.
It operates under explicit instructions to answer only from your documentation, to search before claiming it does not know something, and to decline questions unrelated to your product. If your documentation genuinely does not cover something, it says so and points to support instead of inventing a plausible-sounding answer — which is the failure mode that makes teams distrust these features in the first place.
It respects the same boundaries
Voice is not a side door around your access rules:
- Only published articles are reachable — drafts, hidden articles and trashed content are never read aloud or used as source material.
- A private knowledge base stays private; anonymous callers get nothing.
- Private notes are stripped from both narration and the assistant’s source material.
- Turning the Reader Assistant off switches voice off with it, enforced on the server rather than by hiding the button.
Practical limits worth knowing
A voice conversation runs up to eight minutes, then ends with a message rather than trailing off — long enough for a real question and follow-ups, short enough that a forgotten open tab is not expensive. There is also a daily ceiling on voice conversations per site, which scales with your plan.
Narration and voice conversations are separate features with separate switches, so you can offer one without the other.
Turning it on
The assistant, including voice, lives under Settings → AI (Ved AI) → Enable & Configure. Narration is the Enable read out loud toggle in your article settings. Publish, and your knowledge base can be heard as well as read.