> ## Documentation Index
> Fetch the complete documentation index at: https://miniti.app/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# speech coaching
> training mode - filler words, talk ratio, pace, monologue, questions, clarity
training mode analyzes your delivery in real time. all metrics are computed locally from the transcript - no llm calls, no extra api usage, no extra cost.
switch to training mode with `⌘3` (macos) during a live session or while reviewing a saved meeting.
## metrics
| metric | what it measures |
| ----------------- | -------------------------------------------------------------------------------------------------------------------------------------- |
| filler words | counts of "um", "uh", "like", "basically", etc. - per type, per speaker, per minute. language-specific defaults with custom overrides. |
| talk ratio | how much of the conversation each speaker occupies |
| speaking pace | words per minute per speaker |
| longest monologue | longest uninterrupted stretch (in words) |
| questions asked | count of questions per speaker |
| clarity | average words per turn - shorter turns tend to read as clearer |
## how to use it
training mode is best for coaching yourself on delivery, not grading a meeting overall.
1. switch to training (`⌘3`) during a live session or on a saved meeting
2. focus on **one metric at a time** (fillers first, then talk ratio, etc.)
3. compare similar meeting types (demo vs discovery vs interview) - not everything against one target
4. use the info buttons next to metrics for quick guidance and rough ranges
5. track trends across sessions instead of overreacting to one meeting
## how to read the metrics
* **fillers** - lower is usually better. spikes are normal during brainstorming or complex explanations.
* **talk ratio** - depends on your role. a presenter should be high. an interviewer should be low.
* **pace** - fast is not always better. clarity usually drops when pace gets too high.
* **longest monologue** - useful for spotting when you're not leaving room for questions.
* **questions asked** - directional only. quality matters more than raw count.
* **clarity** - lower average words/turn usually feels easier to follow. lower = clearer = better.
## custom filler words
each language has sensible defaults. to override the list for a language, go to settings → training → filler words.
## real-time nudges
miniti can send a local notification when you've been monologuing or your filler rate spikes - but **only when the app is in the background**, so it doesn't compete with what you can already see in the meeting view. thresholds are fixed: long monologue (\~180 words over 60s), filler spike (>8/min over 60s), 3-minute cooldown per nudge.
nudges are fully local - no llm, no network. enable them in settings → notifications. see [real-time nudges](/docs/features/realtime-nudges) for the full picture.
## notes
* metrics are computed from final transcript segments, not interim
* training applies in both live and saved meeting views
* if a meeting has no transcript content, training sections may be empty
## deep dives
* [filler words](/docs/coaching/filler-words) - per-language lists, customization, how to reduce
* [delivery metrics](/docs/coaching/delivery) - talk ratio, pace, monologue, clarity ranges by meeting type
* [tracking progress](/docs/coaching/tracking-progress) - weekly coaching cycle, comparing meetings, signals of real improvement