Overview
DeepGram is a leading speech-to-text API known for speed and accuracy. Their Nova-3 model delivers state-of-the-art transcription with sub-second processing times. Key Benefits:- ⚡ Ultra-fast - Sub-second transcription processing
- 🎯 High accuracy - Nova-3 achieves industry-leading word error rates
- 🌐 30+ languages - Auto-language detection included
- 🎁 $200 free credits - Generous free tier for new users
Free Tier
What You Get
- $200 in credits upon signup
- No credit card required for free tier
- No expiration on credits (for active accounts)
- All features included - Nova-3, streaming, punctuation, etc.
Getting Your Free API Key
Visit DeepGram Console
Go to console.deepgram.com in your browser.
Sign up for free
Click Sign Up and create an account with:
- Email address
- Password
- Or sign in with Google/GitHub
Verify your email
Check your email inbox for a verification link from DeepGram and click it to confirm your account.
Navigate to API Keys
Once logged in:
- Click on your project name (usually “Default Project”)
- Go to the API Keys section in the left sidebar
Create a new API key
- Click Create New Key
- Give it a name (e.g., “Stenox”)
- Copy the API key - you won’t be able to see it again!
Keep your API key secure! Treat it like a password. Don’t share it publicly or commit it to version control.
Available Models
DeepGram offers several model variants:Nova-3 (Recommended)
Latest and best model:- State-of-the-art accuracy
- Fastest processing (< 1 second)
- Best for: All use cases, default choice
nova-3- Standard, multilingualnova-3-english- Optimized for English
Nova-2
Previous generation:- Still excellent accuracy
- Slightly cheaper than Nova-3
- Best for: Budget-conscious users
nova-2- Standard, multilingualnova-2-english- Optimized for English
Flux
Fast, cost-effective option:- Good accuracy for general use
- Cheaper than Nova models
- Best for: High-volume, non-critical transcription
Language Support
DeepGram supports 30+ languages with auto-detection: Supported Languages:- English (US, UK, Australian, Indian accents)
- Spanish (Spain, Latin America)
- French, German, Italian, Portuguese
- Dutch, Polish, Russian, Turkish
- Chinese (Mandarin), Japanese, Korean, Hindi
- And many more…
- Enable in Stenox profile settings
- DeepGram automatically identifies the language being spoken
- Works across all supported languages
Configuration in Stenox
Configure language
- For English-only: Select
English - For multilingual: Select
Auto-detector specific language
Features
Streaming Transcription
DeepGram uses WebSocket streaming for real-time results:- Audio is sent as you speak
- Transcription begins immediately
- Results return in < 1 second after you stop speaking
Automatic Punctuation
DeepGram automatically adds punctuation:- Periods, commas, question marks
- Capitalization of proper nouns
- Natural sentence structure
Speaker Adaptation
DeepGram adapts to your voice:- Learns your speaking patterns
- Improves accuracy over time
- Better recognition of frequently used terms
Performance
Expected processing times with DeepGram:| Recording Length | Processing Time |
|---|---|
| 5 seconds | < 0.5 seconds |
| 10 seconds | < 1 second |
| 30 seconds | 1-2 seconds |
| 60 seconds | 2-3 seconds |
Usage Tracking
Monitor your credit usage in the DeepGram console:- Visit console.deepgram.com
- Go to Billing in the sidebar
- View:
- Remaining credits ($200 initially)
- Usage this month
- Detailed usage logs
Stenox does not track your DeepGram usage. Check the DeepGram console for accurate credit balance.
Privacy Considerations
What DeepGram sees:- Your audio is sent to DeepGram servers for processing
- DeepGram may log audio and transcripts for quality improvement
- Data is encrypted in transit (HTTPS/WSS)
- Use WhisperKit (local) for 100% private transcription
- Or review DeepGram’s privacy policy and data retention settings
When to Use DeepGram
Maximum accuracy
Nova-3 delivers state-of-the-art transcription quality.
Speed matters
Sub-second processing for near-instant results.
Professional use
Reliable for business emails, documents, and presentations.
Free cloud option
$200 in credits provides generous free usage.
Troubleshooting
API key invalid or not working
API key invalid or not working
- Ensure you copied the entire key (no spaces or extra characters)
- Check that the key is active in DeepGram console
- Try creating a new API key
- Verify your DeepGram account is verified (check email)
Transcription fails or times out
Transcription fails or times out
- Check your internet connection
- Verify you have remaining credits in DeepGram console
- Try a different model (Nova-2 or Flux)
- Check Stenox logs for error messages
Poor accuracy or wrong words
Poor accuracy or wrong words
- Ensure you’re using Nova-3 (best accuracy)
- Check language setting matches your speech
- Reduce background noise
- Speak more clearly or use a better microphone
- Add custom vocabulary in Stenox settings
Running out of credits quickly
Running out of credits quickly
- Check usage in DeepGram console (Billing → Usage)
- Switch to Flux model for cheaper transcription
- Use WhisperKit (local) when privacy isn’t critical to preserve credits
- Create profiles to automatically switch based on use case
Pricing After Free Credits
Once your $200 in credits are used:| Model | Price per Minute | Price per Hour |
|---|---|---|
| Nova-3 | $0.0125 | $0.75 |
| Nova-2 | $0.0100 | $0.60 |
| Flux | $0.0050 | $0.30 |
- 1 hour of dictation per day = ~$22.50/month (Nova-3)
- 30 minutes per day = ~$11.25/month (Nova-3)
Next Steps
Add AI Enhancement
Combine with Google Gemini for grammar correction and formatting.
Create Profiles
Set up different profiles for different use cases.
Free Tier Guide
Maximize your free credits across all providers.
Cloud Setup Guide
Recommended cloud configuration for best results.