Write, compose, and create
5x faster with AI

Absolutely. HyperWhisper can be used entirely offline on your Mac using local AI models - no internet required. For users who want maximum accuracy and speed, we also offer optional cloud processing via various cloud providers.

Yes, it works everywhere you can type. Whether you're composing emails, writing documents, coding, or chatting. Just hit the shortcut, speak, and watch your words appear instantly.

We achieve industry-leading 99% accuracy using state-of-the-art AI models. You can boost accuracy even further by adding your own custom vocabulary for technical terms, names, and specialized jargon.

We never store anything from your transcriptions. When using local mode, no data ever leaves your Mac. When using cloud transcription, your data is sent to the cloud provider you choose (e.g., OpenAI, Groq, Fireworks AI, Deepgram, AssemblyAI, ElevenLabs, or Mistral). You can verify this yourself with a network inspector such as Proxyman.

We believe in simple, fair pricing. Pay once for $39 and use HyperWhisper forever. No monthly fees, no surprises. The free version lets you key features with 3 minutes daily transcription.

Any Mac running macOS 14 or newer works great - both Apple Silicon and Intel chips are fully supported. For optimal performance, 8GB of RAM or more is recommended.

Over 100 languages are supported out of the box, from English and Spanish to Mandarin and Arabic. Most models can automatically detect which language you're speaking and switch seamlessly.

You bring your own API key for the following providers: Fireworks AI, Groq, OpenAI, Deepgram, AssemblyAI, and ElevenLabs. Some providers offer free credits to new users; Deepgram offers $200, AssemblyAI offers $50. This works out significantly cheaper in the long-run than any subscription. You can also use HyperWhisper Cloud (built-in) which requires no API key and is pay-as-you-go. It uses Deepgram for transcription and Groq for post-processing behind the scenes. But remember, the local offline mode works without any API key and is completely free.

HyperWhisper Cloud is a built-in cloud transcription aimed at less technical users who want to get started quickly. It's pay as you go. Buying a license gives you $5 in transcription credits. You can add more credits later. We forward your requests to Deepgram's Nova-3 model for transcription and Groq's Llama model for post-processing. We store nothing besides how many credits you consumed. You pay exactly what we pay to these providers. We charge no markup. For peace of mind, you can view the source code for HyperWhisper Cloud here.

Transcription credits solve a real problem: many users want to use cloud transcription but don't want to deal with making more accounts and setting up API keys. HyperWhisper Cloud lets you transcribe with no extra configuration needed. Credits are the payment system for that convenience. We could charge a subscription instead, but we believe in one-time pricing. So credits are your pay-as-you-go option within that one-time purchase model. If you prefer local transcription (free) or you're comfortable setting up your own API keys (which can be cheaper for high volume), you don't need credits at all.

You can use your license on as many devices as you like, subject to our Fair Usage Policy as outlined in our Terms of Service. This policy is designed to be flexible for legitimate personal and business use while preventing abuse.

For offline transcription, we offer multiple sizes of the Whisper model (made by OpenAI, supporting 99+ languages). We also offer NVidia's Parakeet models (supporting 25+ languages). For offline post-processing, we offer multiple sizes of the Gemma 3 model (made by Google).

Pro users can craft unlimited custom modes tailored to their workflow. Set specific formatting rules, add specialized vocabulary, and define writing styles for any use case you can imagine.

Yes, we offer specialised models for medical and financial domains. These are available through the Deepgram provider (which includes $200 of free credits for new users).

This was made by the YouTuber Ray Amjad.