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

# Providers

> All supported speech-to-text engines — local and cloud — with setup instructions.

Handless supports local (on-device) and cloud-based speech-to-text engines. Local engines need no internet. Cloud providers offer higher accuracy and broader language support but require API keys.

## Local Engines

Download models in **Settings → Models → Library** and run them entirely on your Mac.

| Engine                  | Languages | Speed     | Notes                                                                            |
| ----------------------- | --------- | --------- | -------------------------------------------------------------------------------- |
| **Whisper**             | 99+       | Moderate  | OpenAI's model. Multiple sizes (small → large). Supports translation to English. |
| **Parakeet**            | English   | Fast      | NVIDIA NeMo. Optimized for English with high accuracy.                           |
| **Moonshine**           | English   | Very fast | Ultra-lightweight. Great for quick dictation.                                    |
| **Moonshine Streaming** | English   | Real-time | Streaming variant — shows text as you speak.                                     |
| **SenseVoice**          | Multiple  | Fast      | Multilingual with built-in voice activity detection.                             |

### Model Sizes

Larger models are more accurate but slower and use more memory. Choose based on your needs:

* **Small** (\~40–150 MB) — fastest, good for quick notes
* **Medium** (\~500 MB–1.5 GB) — balanced accuracy and speed
* **Large** (\~1.5–3+ GB) — highest accuracy

## Cloud Providers

Connect your own API keys in **Settings → Models → My Models**.

| Provider            | Default Model          | Streaming | Translation | Diarization |
| ------------------- | ---------------------- | --------- | ----------- | ----------- |
| **OpenAI**          | gpt-4o-mini-transcribe | Yes       | Yes         | No          |
| **Groq**            | whisper-large-v3-turbo | No        | No          | No          |
| **Deepgram**        | nova-3                 | Yes       | No          | Yes         |
| **AssemblyAI**      | universal-3-pro        | Yes       | No          | Yes         |
| **ElevenLabs**      | scribe\_v2             | Yes       | No          | Yes         |
| **Mistral AI**      | voxtral-mini-latest    | Yes       | No          | Yes         |
| **Soniox**          | stt-rt-v4              | Yes       | Yes         | Yes         |
| **Fireworks AI**    | whisper-v3             | Yes       | Yes         | Yes         |
| **Cartesia**        | ink-whisper            | No        | No          | No          |
| **Custom / Ollama** | User-configured        | No        | No          | No          |

### Setting Up a Cloud Provider

1. Open **Settings → Models → My Models**
2. Select a provider from the dropdown
3. Enter your API key (click the console link to get one if needed)
4. Click **Test Connection** to verify
5. Configure provider-specific options (language, temperature, etc.)

### Provider Options

Available options vary by provider. Common ones include:

* **Language** — set a specific language or leave on auto-detect
* **Temperature** — controls randomness (lower = more deterministic)
* **Prompt / Glossary** — provide context to improve accuracy (see [Dictionary](/dictionary))
* **Speaker Diarization** — identify different speakers in the audio
* **Realtime Streaming** — show text incrementally as you speak

## Switching Providers

Switch between providers from the main interface:

1. Click the current provider name in the sidebar
2. Select from your configured cloud providers or downloaded local models

Settings for each provider are saved independently — switching doesn't lose your configuration.