Skip to main content
OpenWhispr offers three transcription modes to balance privacy, speed, cost, and offline capabilities. Choose the mode that best fits your workflow.

​
Overview of Modes

OpenWhispr Cloud

Instant transcription with no API keys needed. Free and Pro plans available.

BYOK (Bring Your Own Key)

Use your own API keys from OpenAI, Groq, or Mistral for full control.

Local Processing

Complete privacy — your voice never leaves your device. Works offline.

​
Mode Comparison

FeatureOpenWhispr CloudBYOKLocal Processing
Privacy⚠️ Audio sent to cloud⚠️ Audio sent to provider✅ Audio stays on device
Speed⚡ Very fast⚡ Fast🐢 Slower (depends on hardware)
CostFree tier or $9/mo ProPay per useFree (after model download)
Offline❌ Requires internet❌ Requires internet✅ Works offline
SetupEmail/Google sign-inAPI key requiredDownload models (75MB - 3GB)
Accuracy⭐⭐⭐⭐⭐ Excellent⭐⭐⭐⭐⭐ Excellent⭐⭐⭐⭐ Very good
Model ChoicePre-selectedMultiple providers6 Whisper models + Parakeet

​
OpenWhispr Cloud

The easiest way to get started — no API keys, no configuration, just sign in and start dictating.

​
Features

  • Free Tier: 2,000 words per week
  • Pro Tier: Unlimited transcriptions for $9/month
  • 7-Day Free Trial: Try Pro features risk-free
  • Sign-In Options: Google OAuth or email/password
  • Managed Infrastructure: No API key management needed

​
When to Use

Getting Started

Perfect for trying OpenWhispr without any setup or configuration.

Casual Users

Ideal if you dictate occasionally and want a simple experience.

High Volume

Pro plan offers unlimited transcriptions for heavy users.

Team Usage

Share referral links to earn free Pro months for your team.

​
How to Set Up

1

Open Settings

Go to Settings → Transcription
2

Select OpenWhispr Cloud

Choose the OpenWhispr Cloud option (should be selected by default if signed in)
3

Sign In

If not already signed in, click Sign In and choose Google or email/password
4

Start Dictating

Your transcriptions are automatically counted toward your weekly quota or Pro plan
Check your usage at any time in Settings → Account to see how many words you’ve used this week.

​
Referral Program

Earn free Pro months by inviting friends and colleagues to OpenWhispr Cloud. How it works:
  1. Go to Settings → Referrals to get your unique referral link
  2. Share your link via email or copy it to share anywhere
  3. When someone signs up using your link and uses 2,000 words, you both earn rewards
  4. Track your referrals and rewards in the Referrals dashboard
Rewards:
  • You earn: 1 free Pro month for each completed referral
  • They earn: Extended free trial or bonus credits
  • No limit on referrals — invite as many people as you want!
A referral is “completed” when the person you invited transcribes at least 2,000 words using OpenWhispr Cloud.

​
BYOK (Bring Your Own Key)

Bring Your Own Key mode lets you use your own API keys from supported providers, giving you full control over billing and model selection.

​
Supported Providers

Models Available:
  • gpt-4o-transcribe — Most accurate
  • gpt-4o-mini-transcribe — Fast and accurate
  • whisper-1 — Original Whisper model
Cost: ~$0.006 per minuteGet API Key: OpenAI Platform

​
When to Use

High Volume

If you transcribe heavily and want to manage your own costs.

Enterprise

Organizations with existing API contracts or custom endpoints.

Speed Priority

Groq’s ultra-fast inference is perfect for real-time workflows.

Custom Infrastructure

You’re running your own Whisper API or using alternative providers.

​
How to Set Up

1

Get API Key

Sign up for your chosen provider and generate an API key
2

Open Settings

Navigate to Settings → Transcription
3

Select BYOK Mode

Choose Bring Your Own Key (BYOK) from the mode selector
4

Choose Provider

Select your provider (OpenAI, Groq, Mistral, or Custom)
5

Enter API Key

Paste your API key into the appropriate field
6

Select Model

Choose your preferred model from the dropdown
7

Save

Your settings are saved automatically
Keep your API keys secure! They are stored locally in your system keychain and never sent to OpenWhispr servers.

​
Local Processing

Local processing keeps your voice data completely private by running speech recognition directly on your device using Whisper or NVIDIA Parakeet models.

​
Available Engines

Overview: OpenAI’s Whisper model via high-performance C++ implementationModels:
  • Tiny (75MB) — Fastest, lower quality
  • Base (142MB) — ⭐ Recommended balance
  • Small (466MB) — Better quality, slower
  • Medium (1.5GB) — High quality
  • Large (3GB) — Best quality, slowest
  • Turbo (1.6GB) — Fast with good quality
Languages: 58 languages supportedHardware: Runs on CPU (no GPU required)

​
When to Use

Privacy-First

Your audio never leaves your device — perfect for sensitive content.

Offline Work

Continue dictating without internet connection (flights, remote areas).

Cost Savings

After downloading models, transcription is completely free.

Custom Control

Choose exactly which model and parameters to use.

​
How to Set Up

1

Open Settings

Navigate to Settings → Transcription
2

Select Local Processing

Toggle Use Local Processing to ON
3

Choose Engine

Select Whisper or NVIDIA Parakeet as your transcription engine
4

Download Model

  • For Whisper: Select model size (Base recommended for most users)
  • For Parakeet: Select Parakeet TDT 0.6B
  • Click Download Model and wait for completion
5

Start Dictating

Your audio is now processed locally on your device

​
Model Storage

  • macOS/Linux: ~/.cache/openwhispr/whisper-models/ or ~/.cache/openwhispr/parakeet-models/
  • Windows: %USERPROFILE%\.cache\openwhispr\whisper-models\ or parakeet-models\
You can remove downloaded models in Settings → Transcription → Local Model Storage → Remove Downloaded Models.
Start with the Base Whisper model (142MB) — it offers the best balance of speed and accuracy for most users.

​
Performance Tips

  • Quick dictation: Use Tiny or Base
  • General use: Use Base or Small (recommended)
  • High accuracy: Use Medium or Large
  • Fast multilingual: Use Parakeet TDT 0.6B
  • Larger models require more RAM (Large needs ~4GB free)
  • Modern CPUs (2018+) handle Base/Small models well
  • Older hardware should stick to Tiny or Base
  • Parakeet is optimized for CPU inference
  • Tiny: ~0.1x real-time (10-second audio = 1 second)
  • Base: ~0.3x real-time (10-second audio = 3 seconds)
  • Small: ~0.8x real-time (10-second audio = 8 seconds)
  • Medium/Large: 1-3x real-time (can be slower than audio length)
  • Parakeet: ~0.2x real-time (very fast)

​
Switching Between Modes

You can switch modes at any time:
  1. Open Settings → Transcription
  2. Select your desired mode
  3. Configure any required API keys or download models
  4. Start dictating — the new mode takes effect immediately
Your mode preference is saved automatically and persists across app restarts.

​
Which Mode Should You Choose?

1

Start with OpenWhispr Cloud

If you’re new to OpenWhispr, begin with the cloud mode — it’s the easiest way to get started and includes a generous free tier.
2

Try Local for Privacy

If you handle sensitive information (medical, legal, confidential), switch to local processing for complete privacy.
3

Use BYOK for Control

If you need specific models, custom endpoints, or want to manage your own costs, use BYOK mode.

​
Next Steps

AI Processing

Enhance your transcriptions with AI formatting and cleanup

Custom Dictionary

Add technical terms and names for better accuracy

Notes System

Create organized voice notes with transcription

Basic Dictation

Learn the fundamentals of using OpenWhispr

Build docs developers (and LLMs) love

Get started for freeTalk to us