HoAh
← Back to Home

Get Started with HoAh AI

HoAh uses powerful AI models to polish and translate your voice. To do this, you need a "Key" from an AI provider. Don't worry, it's easy and often free!

1. What is an API Key?
Think of an API Key as a digital VIP Pass. It allows HoAh to send your text to a super-smart AI brain (like ChatGPT) for processing and get the result back. You are in full control of this key.
2. Choose a Provider

There are many AI providers. We recommended Groq for speed (instant) or OpenAI for quality.

⚑️

Groq

Extremely fast. Free tier available. Perfect for real-time dictation polish.

πŸš€

Cerebras

World's fastest AI inference. Free tier available. Powered by custom wafer-scale chips.

🧠

OpenAI

The industry standard. High intelligence. Paid (pay-as-you-go). Best for complex translations.

πŸ’Ž

Google Gemini

Google's powerful model. Free tier available. Good balance of speed and smarts.

3. Cloud Streaming Dictation

HoAh also supports cloud real-time streaming dictation for lower latency β€” text appears as you speak. We recommend ElevenLabs for the best overall experience. An API key from the provider is required.

πŸŽ™οΈ

ElevenLabs β˜… Recommended

Scribe v2 model. Best overall streaming experience. Excellent multilingual support.

🧠

OpenAI

GPT-4o Transcribe. Fastest streaming. Smart voice activity detection built in.

☁️

Amazon Transcribe

AWS streaming transcription. Enterprise-grade. Reliable and multilingual.

4. Alternative: Run AI Locally with Ollama

Don't want to use cloud APIs? Ollama lets you run AI models entirely on your Mac β€” no API key, no internet, no cost. Perfect for privacy-focused users.

πŸ¦™

Ollama (Local)

Run AI models locally on your Mac. 100% free, 100% private. No API key needed.

Setup in 3 steps:

  1. Install Ollama from ollama.com
  2. Open Terminal and run: ollama pull qwen2.5:1.5b
  3. In HoAh, add an AI Configuration β†’ select Ollama (Local) as the provider

Recommended models:

qwen2.5:1.5b β€” Fast, lightweight (default)
llama3.2:3b β€” Good balance of speed and quality
gemma2:2b β€” Google's compact model
phi3:3.8b β€” Microsoft's efficient model

Tip: Run ollama list to see installed models. If HoAh can't connect, make sure Ollama is running (ollama serve).