Setup guide

From download to your first translated class in about ten minutes. You need a Mac with Apple silicon and macOS 14 or later.

Download for macOS

1. Download and install

  1. Download Liveglot from the button at the top of this page. It's a signed and notarized disk image.
  2. Open the DMG and drag Liveglot into Applications.
  3. Open Liveglot from Applications. The 7-day free trial starts now, with every feature.

Liveglot updates itself: when a new version is out it offers to install it. You can also check in Settings → About.

2. Allow audio recording

To hear what your Mac plays, macOS asks for the Screen & System Audio Recording permission the first time you press Start. Liveglot only records audio, never the screen.

  1. Press Start recording. A small guide appears next to System Settings.
  2. Drag Liveglot into the list, or turn on its switch.
  3. Click Restart. macOS applies the permission only after a restart.

Using the microphone as the source asks for the Microphone permission the same way.

3. Set up speech recognition

Liveglot turns speech into text with one of two engines. The local model is free and private; Soniox is a paid cloud service that is a little faster and more accurate.

Option A — the free local model (recommended). Install whisper.cpp with Homebrew (get Homebrew at brew.sh if you don't have it):

brew install whisper-cpp

Then download a model. large-v3-turbo (q5_0, about 570 MB) is accurate and fast on any Apple silicon Mac:

mkdir -p ~/.cache/whisper.cpp
curl -L -o ~/.cache/whisper.cpp/ggml-large-v3-turbo-q5_0.bin \
  https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-large-v3-turbo-q5_0.bin

On an 8 GB Mac or for lower latency, the small model (about 490 MB) is lighter:

curl -L -o ~/.cache/whisper.cpp/ggml-small.bin \
  https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin
  1. Open Settings → Recognition and choose Local model.
  2. Pick the model from the list. If it isn't there, click Rescan models or Choose model file….
  3. Set Spoken language to the speaker's language (English for most classes) rather than Auto detect: it is faster and more accurate.
  4. Click Save. The first sentence takes a few seconds while the model loads.

Liveglot finds whisper-server on its own after Homebrew installs it. If it says it can't, set its path in Settings → Recognition (usually /opt/homebrew/bin/whisper-server).

Option B — Soniox cloud. Create an API key in the Soniox console (console.soniox.com), then in Settings → Recognition choose Soniox cloud, paste the key, choose the region of your project and click Test connection.

4. Set up translation

Translation and the chat use any OpenAI-compatible service with your own API key. A one-hour class usually costs a few cents with a small model.

  1. Get an API key from the service you choose.
  2. Open Settings → Translate & chat. Fill in Base URL, API key and Model, from the table below.
  3. Choose Translate into, the language you want to read.
  4. Click Test connection, then Save.
ServiceBase URLModelAPI key from
OpenAIhttps://api.openai.com/v1gpt-4o-miniplatform.openai.com
DeepSeekhttps://api.deepseek.com/v1deepseek-chatplatform.deepseek.com
OpenRouterhttps://openrouter.ai/api/v1openai/gpt-4o-miniopenrouter.ai
Ollamahttp://localhost:11434/v1qwen2.5:7b—

To keep everything on your Mac, run a model with Ollama and use the Ollama row, with no API key:

brew install ollama
ollama serve &
ollama pull qwen2.5:7b

You can keep several profiles, for example a cheap one for translation and a stronger one for the chat.

5. Your first session

  1. Play the class or stream on your Mac.
  2. In Liveglot, choose the source: System audio for a stream, Microphone for the room, or both.
  3. Press Start recording. Captions appear, then their translations, sentence by sentence.
  4. Right-click a line to copy it or quote it in the chat, and ask what you missed.

Every session is saved as a Markdown file in ~/Documents/Liveglot (change it in Settings → Records).

6. Activate your license

After the trial, buy a license once for $19.99 on this website. The page after payment shows your license key, and it is emailed to you too.

  1. Open Settings → License.
  2. Paste the key and click Activate.

A key works on two Macs. To move it to another Mac, click Deactivate on this Mac first.

Troubleshooting

ProblemWhat to do
"whisper-server wasn't found"Run brew install whisper-cpp, or set its path in Settings → Recognition.
"No audio arrived"Check that something is playing and the source is right. For system audio, check the permission in System Settings → Privacy & Security → Screen & System Audio Recording.
Recording access is on but still failsIn the permission guide, choose More → Reset recording access, allow it again and restart Liveglot.
Translations are slowUse a smaller, faster model, such as gpt-4o-mini or deepseek-chat.
Recognition is slow or lagsSet Spoken language instead of Auto detect, or use the small model.

Still stuck? Write to [email protected].