personal memory agent
Improve segmentation prompt for timestamp-free transcripts
Add topic-based segmentation mode for transcripts without embedded
timestamps (e.g. chat-format meeting transcripts). The model now
segments by conversation shifts and estimates timestamps from word
count at ~130 wpm, instead of hallucinating uniform line splits.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
+15
-13
+15
-13
think/detect_transcript_segment.md
+15
-13
think/detect_transcript_segment.md
···
4
4
label: Segmentation
5
5
group: Import
6
6
---
7
-
You are a transcript analyzer that identifies 5-minute segment boundaries with absolute timestamps.
7
+
You are a transcript analyzer that splits transcripts into ~5-minute segments.
8
8
9
-
TASK: Find ~5-minute segment boundaries and return their line numbers with absolute time-of-day timestamps.
9
+
TASK: Find segment boundaries and return their line numbers with absolute time-of-day timestamps.
10
10
11
11
INPUT FORMAT:
12
12
- First line: "START_TIME: HH:MM:SS" - the absolute start time of this transcript
13
13
- Remaining lines: Transcript with line numbers prepended as "N: content"
14
-
- Timestamps in transcript may be relative (00:00:00, 05:30) or absolute (14:30:22)
15
14
16
15
OUTPUT FORMAT:
17
16
- JSON array of objects with "start_at" and "line" fields
···
19
18
- "line": Line number where this segment begins
20
19
- Example: [{"start_at":"12:00:00","line":1},{"start_at":"12:05:23","line":42}]
21
20
22
-
REQUIREMENTS:
23
-
1. First segment starts at the provided START_TIME on line 1
24
-
2. Detect if transcript uses relative or absolute timestamps:
25
-
- Relative (00:00:00, 05:30): Add to START_TIME to get absolute
26
-
- Absolute (14:30:22): Use directly
27
-
3. Find boundaries near 5-minute intervals from start
28
-
4. Output all times as absolute HH:MM:SS
21
+
SEGMENTATION MODES:
29
22
30
-
EDGE CASES:
31
-
- Transcript < 5 minutes: return single segment at START_TIME
32
-
- No valid timestamps: return single segment at START_TIME
23
+
1. **Timestamped transcripts** — if the text contains timestamps (relative like 00:05:30 or absolute like 14:30:22), use them to find boundaries near 5-minute intervals. Convert relative timestamps by adding to START_TIME.
24
+
25
+
2. **Timestamp-free transcripts** — if the text has NO timestamps (e.g. just speaker labels and dialogue), segment by **topic and conversation shifts** instead:
26
+
- Find natural break points where the conversation changes subject
27
+
- Estimate time from position: assume ~130 words/minute speaking rate, calculate total duration from word count, then assign proportional timestamps from START_TIME
28
+
- Aim for segments roughly 5 minutes of estimated speaking time, but prioritize clean topic breaks over exact intervals
29
+
- NEVER distribute lines uniformly — segments should vary in size based on where topics actually change
30
+
31
+
REQUIREMENTS:
32
+
1. First segment always starts at START_TIME on line 1
33
+
2. All output times must be absolute HH:MM:SS
34
+
3. Every transcript gets multiple segments unless it is extremely short (under ~2 minutes estimated)
33
35
34
36
RESPONSE: Return only the JSON array, no additional text.