Diagnosis — how to use it in the Studio
The Diagnosis reads a conversation that is already transcribed or extracted and returns the diagnosis with an HTML report. It does not record or transcribe: that is done in the Input tab.
The best path: take care of the Input
The report inherits the quality of the conversation you loaded. Do it like this:
- In the Input tab, transcribe the audio (Audio card) or extract the WhatsApp conversation (Conversation extraction card) — in WhatsApp, choose INCLUDE MEDIA so the audio, stickers and emojis come along.
- At the end of the Input, check the dialogue and save the HTML report — it is the same material the Diagnosis will use.
- Go to the Diagnosis tab and click LOAD FROM INPUT: it fills in the conversation, the duration, the participants and the events automatically (do not type by hand).
- Check the participants and the duration — that is what the AI uses to score.
- Choose the Language and the Response language, click RUN DIAGNOSIS and open or download the HTML report.
Step by step in the Studio
- Transcribe or extract in the Input tab.
- Open the Diagnosis tab and click LOAD FROM INPUT.
- Check the fields and click RUN DIAGNOSIS.
- Open and download the HTML report — keep it as evidence.
What you see in the form
Language of the conversationLanguage of the conversation. Filled in by the Input.
Response languageLanguage of the report. If you leave it blank, it follows the conversation language.
ParticipantsWho speaks in the conversation. Comes from the Input and only exists in a diarized conversation (with identified speakers).
Conversation durationDuration in seconds. Also only in a diarized conversation.
Call directionIncoming or outgoing call.
Diarized conversationThe dialogue with speaker and time on each turn: Speaker 1: [00:00:00.540 - 00:00:01.139] text.
ConversationAlternative to the diarized conversation: plain text, without speakers or times.
Audio eventsNoises and tags that left the dialogue (e.g. [laughter]).
ReferenceYour own name to find the result later.
What it costs
- The analysis is charged per character of the conversation — the longer the dialogue, the more credits.
- Only valid content counts: labels, times and audio events are not counted.
- Running again charges again: adjust the conversation before running.
- The usage stays in your credit history (see the Dashboard).
If something goes wrong
- Missing data? Use LOAD FROM INPUT — it fills in the conversation, the duration, the participants and the events.
- Participants, duration or events do not show up: they only exist in a diarized conversation (with speakers). In plain text they are ignored.
- No credits: check the balance on the Dashboard.
- The HTML report comes in Portuguese, Spanish and English.
- The result came out empty: check that the conversation was really loaded (in the Input the dialogue needs turns).
The Diagnosis is analytical (quality and compliance) and does not replace the Risk Audit. You are responsible for having the consent of the people in the conversation (GDPR/LGPD).