> ## Documentation Index
> Fetch the complete documentation index at: https://superwhisper.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Slow dictations or heavy memory use

> Reduce slow dictation processing and high memory use in Superwhisper. Model choice and model load time usually matter most.

## Overview

If dictations process slowly or the app feels heavy, the tips below cover the areas with the biggest impact. Four factors drive performance:

> [card] System Resources
>
> Processing power and available memory matter most with local models.

> [card] Internet Connection
>
> Cloud processing depends on your connection speed and stability.

> [card] Model Selection
>
> Models differ in processing time and memory use. Trying a different model is a quick test.

> [card] Processing Type
>
> Voice transcription and AI processing are separate steps with separate costs.

***

## Local vs. cloud models

### Local models

Local models run directly on your computer, so your hardware sets the processing speed.

**1. Tune Voice model active duration**

![Superwhisper Voice model active duration setting](https://mintcdn.com/superwhisper/SFDRA4G70egO6eAv/images/screens/mac/settings-voice-model.png?fit=max\&auto=format\&n=SFDRA4G70egO6eAv\&q=85\&s=22239e236f06192cb3deb51a642c2822)

In **Settings → Configuration → Advanced settings**, under **Voice model**, adjust **Voice model active duration**. See [advanced settings](https://superwhisper.com/docs/get-started/settings-advanced.md#voice-model). It sets how long voice models stay loaded:

* Settings range from 10 seconds to 1 hour
* Shorter durations free memory but reload the model for each new dictation
* Longer durations keep the model ready but hold more memory

**2. Match the model to your hardware**

* Local models need free memory; close memory-heavy applications while dictating
* If your machine still struggles, pick a smaller model such as **Whisper Base**, or switch that mode to a cloud model

### Cloud models

Cloud models process your dictation on remote servers, so your hardware matters less. They need a stable internet connection. On a slow or unstable connection, a local model can work better.

***

## Speed up processing

**1. Skip AI processing when you don't need it**

Transcription and AI processing are separate steps. The [Voice to Text](https://superwhisper.com/docs/modes/built-in.md#voice-to-text) mode skips AI processing entirely, so only the transcription step runs. Use it when you don't need formatting.

**2. Try a different model**

* Cloud voice models, such as **S1-Voice**, move the work off your machine
* Local models, such as **Parakeet Multilanguage** (494 MB), run offline and keep audio on your device
* Smaller local models, such as **Whisper Tiny** or **Whisper Base**, use less memory

**3. Find the slow step in History**

Open a dictation in [History](https://superwhisper.com/docs/get-started/history.md) and compare the voice and AI processing times in the right sidebar. Whichever step is slow, switch that model.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.