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

# GPU Acceleration

> Enable hardware acceleration for 5-10x faster transcription on Windows

VoiceTypr uses hardware acceleration to dramatically speed up transcription. GPU acceleration can make transcription **5-10x faster** than CPU-only processing.

<Info>
  GPU acceleration is **automatic** - VoiceTypr detects and uses your GPU if available. No manual configuration required.
</Info>

## Windows GPU Support

VoiceTypr supports three major GPU vendors on Windows:

<CardGroup cols={3}>
  <Card title="NVIDIA" icon="n" href="https://www.nvidia.com/drivers">
    GeForce, RTX, Quadro

    **Most common**

    Best performance
  </Card>

  <Card title="AMD" icon="a" href="https://www.amd.com/support">
    Radeon series

    **Good support**

    Solid performance
  </Card>

  <Card title="Intel" icon="i" href="https://www.intel.com/content/www/us/en/support/products/80939/graphics.html">
    Iris Xe, Arc series

    **Newer GPUs**

    Good for laptops
  </Card>
</CardGroup>

### How It Works

1. **Automatic Detection**: VoiceTypr checks for compatible GPUs on startup
2. **Automatic Selection**: If a GPU is found, it's used automatically
3. **CPU Fallback**: If no GPU is available, CPU processing is used

<Warning>
  No user intervention needed - VoiceTypr handles everything automatically.
</Warning>

## Driver Requirements

Up-to-date graphics drivers are **critical** for GPU acceleration:

### NVIDIA Drivers

<Steps>
  <Step title="Check Current Version">
    Right-click desktop → NVIDIA Control Panel → Help → System Information
  </Step>

  <Step title="Download Latest Drivers">
    Visit [NVIDIA Drivers](https://www.nvidia.com/drivers)

    Select your GPU model
  </Step>

  <Step title="Install and Restart">
    Run the installer

    Restart your computer
  </Step>
</Steps>

**Recommended versions:**

* GeForce Game Ready Driver 500+
* Studio Driver 500+

### AMD Drivers

<Steps>
  <Step title="Check Current Version">
    Right-click desktop → AMD Software → Settings → System
  </Step>

  <Step title="Download Latest Drivers">
    Visit [AMD Support](https://www.amd.com/support)

    Auto-detect your GPU or select manually
  </Step>

  <Step title="Install and Restart">
    Run AMD Software Installer

    Restart your computer
  </Step>
</Steps>

**Recommended versions:**

* Adrenalin 22.0+
* PRO Driver 22.0+

### Intel Drivers

<Steps>
  <Step title="Check Current Version">
    Right-click desktop → Intel Graphics Command Center → System
  </Step>

  <Step title="Download Latest Drivers">
    Visit [Intel Graphics Drivers](https://www.intel.com/content/www/us/en/support/products/80939/graphics.html)

    Use Driver & Support Assistant for auto-detection
  </Step>

  <Step title="Install and Restart">
    Run the installer

    Restart your computer
  </Step>
</Steps>

**Recommended versions:**

* Intel Arc Graphics Driver 30.0+
* Intel Iris Xe Graphics Driver 30.0+

<Tip>
  Enable Windows Update to automatically receive driver updates for Intel integrated graphics.
</Tip>

## macOS Metal Acceleration

All Macs include automatic hardware acceleration:

### Metal GPU Acceleration

* **All Macs**: Whisper models use Metal GPU acceleration
* **Automatic**: No configuration needed
* **Performance**: 3-5x faster than CPU on Apple Silicon, 2-3x on Intel

### Apple Neural Engine

* **Apple Silicon only**: Parakeet models use the Neural Engine
* **Ultra-fast**: Optimized specifically for Apple Silicon
* **Automatic**: Activated when Parakeet models are selected

<Info>
  macOS users always have hardware acceleration - Metal is built into every Mac.
</Info>

## Performance Expectations

### With GPU Acceleration (Windows)

| Model  | CPU Time | GPU Time | Speedup       |
| ------ | -------- | -------- | ------------- |
| Tiny   | 2.5s     | 0.3s     | **8x faster** |
| Base   | 4s       | 0.5s     | **8x faster** |
| Small  | 8s       | 1s       | **8x faster** |
| Medium | 18s      | 2.5s     | **7x faster** |
| Large  | 35s      | 5s       | **7x faster** |

<Note>
  Times shown for 30-second audio recording. Actual performance varies by GPU model.
</Note>

### Metal Performance (macOS)

| Mac Type                     | Model    | Speedup vs CPU |
| ---------------------------- | -------- | -------------- |
| **Apple Silicon** (M1/M2/M3) | Whisper  | 4-5x faster    |
| **Apple Silicon**            | Parakeet | 6-8x faster    |
| **Intel Mac**                | Whisper  | 2-3x faster    |

## CPU Fallback

If GPU acceleration is unavailable, VoiceTypr automatically uses CPU:

<AccordionGroup>
  <Accordion title="When CPU is Used" icon="microchip" defaultOpen>
    * No compatible GPU detected
    * Graphics drivers are outdated or missing
    * GPU is busy with other tasks
    * GPU memory is insufficient
  </Accordion>

  <Accordion title="CPU Performance" icon="gauge">
    Still fully functional, just slower:

    * Tiny/Base: Acceptable for short recordings
    * Small: Noticeable delay on longer recordings
    * Medium/Large: Can take 30-60 seconds for transcription
  </Accordion>

  <Accordion title="Improving CPU Performance" icon="arrow-up">
    * Use smaller models (Tiny, Base, Small)
    * Close other applications during transcription
    * Upgrade to a system with GPU support
  </Accordion>
</AccordionGroup>

<Warning>
  VoiceTypr **always works** - it automatically falls back to CPU if GPU is unavailable. You'll never see an error.
</Warning>

## Troubleshooting GPU Issues

<AccordionGroup>
  <Accordion title="Slow Transcription" icon="hourglass">
    **Symptoms:** Transcription takes longer than expected

    **Solutions:**

    1. Update graphics drivers (see links above)
    2. Check if GPU is being used by other apps (games, video editing)
    3. Restart VoiceTypr to re-detect GPU
    4. Verify GPU is enabled in Device Manager (Windows)
  </Accordion>

  <Accordion title="GPU Not Detected" icon="search">
    **Symptoms:** CPU fallback being used despite having a GPU

    **Solutions:**

    1. Update graphics drivers (most common fix)
    2. Check GPU is enabled in BIOS/UEFI
    3. Verify GPU appears in Device Manager (Windows)
    4. Ensure GPU has sufficient memory (2GB+ recommended)
  </Accordion>

  <Accordion title="Driver Installation Fails" icon="xmark">
    **Symptoms:** Can't install or update graphics drivers

    **Solutions:**

    1. Uninstall old drivers completely first (use DDU tool)
    2. Download drivers directly from manufacturer (not Windows Update)
    3. Disable antivirus temporarily during installation
    4. Check Windows Update for chipset driver updates
  </Accordion>
</AccordionGroup>

## Verifying GPU Acceleration

How to confirm GPU is being used:

### Windows

<Steps>
  <Step title="Open Task Manager">
    Press `Ctrl+Shift+Esc` or right-click taskbar
  </Step>

  <Step title="Go to Performance Tab">
    Select GPU section
  </Step>

  <Step title="Start Transcription">
    Trigger a recording in VoiceTypr
  </Step>

  <Step title="Watch GPU Usage">
    GPU usage should spike during transcription

    Look for "CUDA" or "Compute" activity
  </Step>
</Steps>

### macOS

<Steps>
  <Step title="Open Activity Monitor">
    Applications → Utilities → Activity Monitor
  </Step>

  <Step title="Select GPU History">
    Window → GPU History
  </Step>

  <Step title="Start Transcription">
    Trigger a recording in VoiceTypr
  </Step>

  <Step title="Watch GPU Usage">
    GPU usage should increase during transcription
  </Step>
</Steps>

## Best Practices

<Steps>
  <Step title="Keep Drivers Updated">
    Update graphics drivers every 2-3 months

    Enable automatic updates if available
  </Step>

  <Step title="Close Heavy Apps">
    Pause games or video rendering during transcription

    Ensures GPU resources are available
  </Step>

  <Step title="Use Larger Models">
    With GPU acceleration, you can use Large models without performance penalty

    Better accuracy at similar speeds to CPU+Small
  </Step>

  <Step title="Monitor Temperature">
    Ensure GPU cooling is adequate

    Overheating can reduce performance
  </Step>
</Steps>

<Tip>
  With GPU acceleration enabled, the **Large** model transcribes faster than the **Tiny** model on CPU. Take advantage of better accuracy!
</Tip>