Cloud Services
Set up cloud transcription providers like Deepgram, Groq, ElevenLabs, Gemini, and Mistral.
Cloud transcription services process your audio on external servers. They are fast, highly accurate, and do not consume your Mac's processing power. The trade-off is that your audio is sent over the internet, and each provider requires an API key.
Available Cloud Providers
| Provider | Model | Accuracy | Multilingual | Notes |
|---|---|---|---|---|
| Deepgram | Nova | Very good | Yes | Fast and cost-effective. Uses Nova-3 for English, Nova-2 for other languages. |
| Deepgram | Nova-3 Medical | Excellent | English only | Specialized for clinical and medical terminology. |
| Groq | Whisper Large v3 Turbo | Excellent | Yes | Runs Whisper on Groq's fast inference hardware. |
| ElevenLabs | Scribe v2 | Excellent | Yes | #1 ranked STT model (2.3% WER). |
| Gemini | Gemini 3.1 Pro (Preview) | Excellent | Yes | Google's latest model with advanced reasoning capabilities. |
| Gemini | Gemini 3 Flash (Preview) | Very good | Yes | High performance at lower cost. |
| Mistral | Voxtral Mini | Excellent | Yes | Mistral's state-of-the-art transcription model. |
Setting Up a Cloud Provider
Each cloud provider needs an API key before you can use it. The setup process is the same for all providers:
- Open Settings and go to the Transcription tab.
- Select the Cloud filter tab to see all cloud models.
- Find the provider you want. It will show a Setup Required badge.
- Click Configure on the model card.
- Enter your API key in the API Key Configuration field.
- Click Verify. Echo will test the key with the provider.
- If verification succeeds, the card will show a Configured badge.
- Click Set as Default to use this model for transcription.
Where to Get API Keys
Each provider has a dashboard where you can create an API key:
| Provider | Where to Get Your Key |
|---|---|
| Deepgram | Deepgram Console |
| Groq | Groq Console |
| ElevenLabs | ElevenLabs Dashboard |
| Gemini | Google AI Studio |
| Mistral | Mistral Console |
Most providers offer a free tier or trial credits, so you can try a service before committing.
Removing an API Key
If you need to remove a configured API key:
- In the Transcription tab, find the configured cloud model.
- Click the three-dot menu on the model card.
- Select Remove API Key.
This will clear the stored key. If this model was your default, you will need to select a different model.
How Cloud Transcription Works
When you record with a cloud model selected, Echo:
- Captures your audio locally.
- Sends the audio file to the cloud provider's API.
- Receives the transcribed text back.
- Applies your output settings (word replacements, formatting, AI enhancement) locally.
The audio is sent directly to the provider you have selected. Echo does not store or relay your audio through any intermediary servers.
ElevenLabs Scribe V2
Scribe V2 is ElevenLabs' latest speech-to-text model and the #1 ranked STT model on the Artificial Analysis benchmark with a 2.3% word error rate. It supports 90+ languages.
Tips
- Cloud services require an internet connection. If you need to transcribe offline, use a local model instead.
- You can have multiple cloud providers configured at the same time. Only the one set as your default will be used for transcription, but you can switch between them quickly.
- If you see an "API key is missing" error, check that you have entered and verified your key for the provider you are trying to use.
Related
- Transcription Overview -- How the three transcription approaches compare
- Local Models -- Models that run on your Mac without an internet connection
- Language Selection -- Choosing the transcription language
- Personal Vocabulary -- Adding custom words to improve accuracy