After an interview wraps up, Mato processes the recording into scored segments you can review, filter, and select before generating an episode. This guide covers the segment editor, AI scoring, the include/exclude toggle, the transcript view, and how to download a formatted transcript.
How segment processing works
When an interview session ends, a background job runs automatically. It analyzes the transcript, groups the conversation into topically-coherent segments, and scores each one across four dimensions. You do not need to trigger this yourself.
Processing typically takes 1 to 2 minutes. While it runs, the edit page shows a spinner and a status message. If something goes wrong, an error state appears with a retry option.
The four scoring dimensions are:
- Topic relevance (0 to 100): how closely the segment relates to the interview's main topic
- Purpose alignment (0 to 100): how well it fulfills the stated goals of the interview
- Audience value (0 to 100): how interesting or useful the content is for listeners
- Brand fit (0 to 100): how well it matches your podcast's tone and style
These four scores are combined into a single weighted score displayed on each segment card. Segments scoring 60 or above are pre-selected for inclusion.
Open the segment editor
From your podcast's interview list, click an interview that has a Completed status. On the interview detail page, click Edit transcript near the bottom of the page. This opens the segment editor.
If segments are still processing, you will see a loading spinner instead. Come back in a minute or refresh the page.
Read a segment card
Each segment card shows:
- Time range: the start and end timestamps formatted as elapsed time (for example,
1:05or12:30), representing how far into the interview the segment occurs - Topic tags: short labels describing the subjects discussed in that segment
- Combined score: a single number (0 to 100) in the top-right corner
- Text preview: the full transcript text of the segment
- Metadata row: tone (excited, thoughtful, humorous, serious, passionate, or neutral), energy level (low, medium, or high), and primary speaker (host or subject)
Score tiers and color coding
Segment cards are color-coded by their combined score:
- Green border (80 and above): strong content, highly relevant
- Yellow border (60 to 79): solid content worth including
- Red border (below 60): lower-relevance content you may want to skip
These colors give you a quick visual scan of segment quality without reading every score.
Include or exclude segments
Every segment card has a toggle switch in the top-right area, next to the score. The toggle shows either Included (green, with a checkmark) or Excluded (gray, with an X).
Click the toggle to flip a segment's state. Excluded segments dim to 50% opacity and turn grayscale, making them easy to distinguish at a glance.
Segments with a combined score of 60 or higher start as included. Segments below 60 start as excluded. You can override either direction.
Your toggle choices apply within the current session. To confirm which segments make it into the final episode, use the segment picker inside the episode generation wizard (described below).
Review segments during episode generation
You can also review and toggle segments inside the episode generation wizard. When you click Create Episode (or Regenerate) from the interview detail page, a three-step wizard opens:
- Review interview: confirms guest name, topic, purpose, segment count, and average score
- Select segments: shows all segments in a scrollable list with compact cards and checkboxes
- Generate episode: tracks progress as Mato builds the episode from your selection
In step two, three quick-action buttons sit above the segment list:
- Select Recommended: resets selection to segments scoring 60+
- Select All: includes every segment
- Clear All: deselects everything
This is the same data as the edit page, displayed in a more compact format designed for fast selection.
Read the interview transcript
The interview detail page includes a full transcript card below the invite details. Each turn in the transcript displays:
- The speaker name (AI Host or the guest's name)
- An elapsed timestamp showing how far into the interview that turn occurred (for example,
0:24,3:15, or1:02:47) - The spoken text
Timestamps are relative to the session start time. They use a compact format: M:SS for times under ten minutes (like 1:05), MM:SS for times under an hour (like 12:30), and H:MM:SS for longer interviews.
A search bar at the top of the transcript lets you filter turns by keyword. The count updates as you type, showing how many turns match out of the total.
At the bottom of the transcript, a stats line shows the total turn count and a breakdown of host versus guest turns.
Download the transcript
Click the Download button in the transcript controls (top-right of the transcript card) to save a .txt file. The downloaded file formats each turn with its elapsed timestamp in brackets, followed by the speaker name and the text:
[0:24] [Host] Welcome to the show. Tell us about your background. [0:38] [Jane Smith] Thanks for having me. I've been working in renewable energy for about twelve years...
Timestamps in the downloaded file follow the same relative format shown in the dashboard.
Guest review page
Guests receive a link to a public review page where they can read the transcript, leave feedback, request edits, and approve the conversation for publication. This page shows the same elapsed timestamps on each turn and includes a Download transcript (.txt) button in the footer.
The guest review page has four sections:
- Hero: guest name, topic, recording date, duration, and turn count
- Transcript: the full conversation with speaker labels and timestamps
- Feedback: optional fields for general notes and specific edit requests (with a hint that timestamps help)
- Approval: a one-time button to approve the transcript for publication
Once a guest approves, the approval badge shows a green checkmark and the approval date. This status is visible on the interview detail page in your dashboard under Approval Status.
Segment boundaries
Segment boundaries are set by Mato's AI analysis during processing. The system groups transcript turns into topically-coherent sections following these rules:
- A host question stays in the same segment as the guest's answer (never split across segments)
- Segments typically span 1 to 5 minutes of conversation
- Greetings and introductions get their own segment
- Topic changes (signaled by phrases like "moving on to" or a shift in subject matter) trigger a new segment
You cannot manually drag segment boundaries. The include/exclude toggle is the primary way to control which content reaches your episode. When you generate an episode from the interview, Mato trims the selected segments to remove dead air and off-topic tangents at the edges. If a segment still contains content you want to cut, the episode script editor (available after generation) gives you line-by-line editing control.
What to do next
- Edit your episode script to refine the generated dialogue before rendering audio
- Generate and edit your first episode for the full generation workflow