Audio and video

A meeting recording, a company presentation video, an internal course: if someone said it out loud, it can become a source for the bot.

How it works

You upload the video where the owner presents the company services. The system transcribes the audio and indexes the transcription as a document: it is divided into blocks and made searchable. If a customer asks "do you also do installations?" and it is mentioned in the video, the bot answers yes with the details.

What you need to do

1. Upload the audio or video file to the knowledge base.
2. Wait for indexing: long files take more time than a document.
3. Verify with a few questions about what is said.
4. If an important part was transcribed incorrectly, fix it in the extracted text or add a FAQ to clarify it.

Limits

• Only what is said counts: what is seen on screen without being spoken (slides, signs) is not included in the transcription.
• Noisy audio, overlapping voices and strong dialect accents worsen the transcription.
• A one hour chat contains many tangents: for key topics, it is better to integrate with a well written FAQ.

Frequent problems

• The bot misquotes a technical term from the video → it is almost always a transcription error: fix the extracted text or cover the term with a FAQ.

From the app

Not available from the app: upload audio and video from the panel.