# Choose Cloud, your own API keys or on-device processing

Source: https://help.contextli.com/articles/choose-cloud-api-keys-or-on-device-processing

Open **Settings** > **Data Control** and pick where your data is processed: **Contextli Cloud**, **Your API Keys** or **Your Device**. You choose separately for **Transcription** (turning speech into text) and **AI Processing** (rewriting text with contexts and Polished Transcription).

## Change the setting

1. Open Contextli and click **Settings**.
2. Click the **Data Control** tab.
3. Under **Transcription**, click the card you want.
4. Under **AI Processing**, click the card you want.

![Settings, Data Control tab: Contextli Cloud, Your API Keys and Your Device for transcription and AI processing](https://cdn.kamaan.io/kamaan/sites/6ae9f572170cd8d9857a277c/images/contextli-settings-data-control.webp)

## The three options

| Option | What it means | Availability |
|---|---|---|
| **Contextli Cloud** (Recommended) | Fastest and most capable. Processed on our servers, then deleted. Uses your credits. | All plans |
| **Your API Keys** | Data goes to your own provider, not to Contextli. Transcription: Deepgram or OpenAI Whisper key. AI: OpenAI or Anthropic key. | Lifetime plans only |
| **Your Device** | 100% offline. Nothing leaves your computer. | Transcription: paid plans. AI Processing: Pro and Pro Plus |

For plan details see https://contextli.com/pricing.

## Set up your own API keys

1. Under **Transcription** or **AI Processing**, click **Your API Keys**.
2. Choose the provider and paste your API key from that provider's dashboard.
3. Dictate once to test it.

Dictation with your own keys never stops because of Contextli credits; usage is billed by your provider.

## Set up on-device transcription

1. Under **Transcription**, click **Your Device**.
2. Pick a preset: **Fast** (works on most computers), **Balanced** (recommended for everyday use) or **Quality** (for powerful computers).
3. Wait for the model to download. After that, transcription runs without internet.

## Set up on-device AI processing

1. Install Ollama or LM Studio on your computer and make sure it is running.
2. Under **AI Processing**, click **Your Device** and pick a model.

If the card shows **SETUP REQUIRED**, the local AI app is not installed or not running yet.

## Good to know

- On-device processing is slower than Contextli Cloud, especially on older computers.
- The Android app has its own **Offline Dictation** switch in its settings.

## Related articles

- [What happens to my audio and text?](/articles/what-data-does-contextli-see)
- [How credits work](/articles/how-credits-work)
- [Redeem a lifetime deal or AppSumo code](/articles/redeem-a-lifetime-deal-code)
- [Delete your account and data](/articles/delete-your-account-and-data)