granola-performance-tuning

Optimize Granola transcription accuracy, note quality, and processing speed. Use when improving transcription quality, reducing processing time, optimizing templates for better AI output, or tuning audio setup. Trigger: "granola performance", "granola accuracy", "granola quality", "improve granola", "granola transcription better".

Zur Installation springen

Repository: jeremylongshore/tons-of-skills-marketplace
Letzte Quellaktivität: 19. August 2026 um 12:13
Erkannte Sprache von SKILL.md: Englisch
Sterne: 2.677
Forks: 392

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

Zurück zur Suche Zurück zur Installation

name	granola-performance-tuning
description	Optimize Granola transcription accuracy, note quality, and processing speed. Use when improving transcription quality, reducing processing time, optimizing templates for better AI output, or tuning audio setup. Trigger: "granola performance", "granola accuracy", "granola quality", "improve granola", "granola transcription better".
allowed-tools	Read, Write, Edit
version	1.13.0
license	MIT
author	Jeremy Longshore <jeremy@intentsolutions.io>
tags	["saas","granola","performance","transcription"]
compatibility	Designed for Claude Code

Granola Performance Tuning

Overview

Optimize Granola output quality across three dimensions: audio/transcription accuracy, AI enhancement quality, and integration speed. Granola's AI (GPT-4o/Claude) produces better output when it has clean audio, well-typed notes, and structured templates.

Prerequisites

Working Granola installation with meetings captured
Willingness to improve audio setup and meeting practices
At least 3-5 meetings captured to establish baseline quality

Instructions

Step 1 — Optimize Audio for Transcription

Granola captures system audio from your device. Transcription accuracy depends entirely on audio quality:

Hardware recommendations (by priority):

Setup	Accuracy Impact	Recommendation
Wired headset with mic	Highest	Best for solo/remote meetings
USB condenser mic	High	Best for in-office, multiple speakers
Laptop built-in mic	Medium	Acceptable for quiet environments
Bluetooth headset	Variable	May cause dropouts — test first
Speakerphone in room	Low	Echo and distance degrade accuracy

Audio configuration checklist:

Correct input device selected in System Settings > Sound > Input
Input volume at 75-100% (not too low, not clipping)
Audio enhancements disabled (Windows: right-click device > Properties > disable enhancements)
No conflicting virtual audio software (Loopback, BlackHole, etc.)
Bluetooth device stable (or switch to wired if experiencing drops)

Room setup:

Minimal background noise (close doors, turn off fans)
Soft surfaces to reduce echo (avoid glass-walled conference rooms)
Mic within 12 inches of speaker(s)

Practice	Impact	Why It Helps
State names when assigning work	High	"Sarah, can you handle the API spec?" enables correct attribution
Use explicit action language	High	"Action item: review by Friday" — AI detects structured language
One speaker at a time	High	Crosstalk confuses speaker diarization
Summarize decisions verbally	Medium	"So we've decided to go with option B" — AI captures decisions
Spell technical terms first time	Medium	"We'll use Kubernetes, K-U-B-E-R-N-E-T-E-S" — improves accuracy
Type notes during the meeting	High	Your notes give the AI critical context for enhancement
Brief recap at meeting end	Medium	"To summarize, we agreed on X, Y, and Z" — improves summary

## Summary
[2-3 sentence overview of the meeting]

## Key Decisions
[Bullet list of decisions made, with reasoning]

## Action Items
[Format: - [ ] @person: task (due date)]

## Open Questions
[Items that need follow-up or weren't resolved]

## Next Steps
[What happens after this meeting]

"What did Mike say about the timeline?"
→ Searches transcript for Mike's statements about timeline

"Were there any disagreements that aren't captured in the summary?"
→ Analyzes transcript for conflicting viewpoints

"Add the budget numbers that were discussed"
→ Pulls specific figures from the transcript

"Rewrite the action items with more detail"
→ Expands terse action items with transcript context

Metric	Target	How to Measure
Transcription accuracy	>95% word accuracy	Spot-check 2-3 min of transcript vs. audio
Action item detection	>90% captured	Compare enhanced notes to manual list
Decision accuracy	100% correct	Verify all listed decisions actually happened
Processing time	<2 min for 30-min meeting	Timestamp when meeting ends vs. when notes are ready
Enhancement usefulness	4+/5 team rating	Monthly survey: "How useful are Granola notes?"

Issue	Cause	Fix
<85% transcription accuracy	Poor microphone or noisy room	Upgrade to wired headset, reduce background noise
Action items missed	Vague language ("someone should...")	Use explicit format: "Action item: @person does X by Y"
Wrong speaker attribution	Crosstalk or no name usage	State names, avoid overlapping speech
Slow processing (>5 min)	Long meeting or server load	Normal for 2+ hour meetings; check status.granola.ai
Hallucinated decisions	AI filling template sections	Review before sharing; remove decisions that didn't happen