Download the Optimized Core ML Model

Download the transcription model

Sequence uses a Core ML conversion of NVIDIA’s Parakeet v3 for on-device speech recognition. Read why Sequence uses Core ML for the engineering background.

1. Sign in and activate a trial or licence.
2. Open Settings → Transcription.
3. In the Transcription model group, review the download size shown on the button, then start the download.
4. Wait for downloading and verification to finish before starting a transcription.

The app obtains the model through your activated account. Download size depends on the current model pack; use the displayed size rather than a fixed estimate. Downloading requires an internet connection.

Local transcription

After the model is installed, transcription runs locally on your Mac. Your media is not uploaded to a cloud transcription service. The resulting transcript is stored in your local encrypted library database.

You can transcribe and edit media without installing the separate speaker model.

First run and warm runs

The first transcription can take longer while Sequence prepares the model for your Mac. Cached preparation helps later runs.

Sequence keeps the model loaded briefly after transcription so nearby jobs can reuse it, then unloads it when idle to release memory. Unloading from memory does not delete the downloaded model.

Manage the installed model

Settings → Transcription shows the installation status and model path. Click the path to reveal the model in Finder. Use the trash button and confirm deletion to remove the downloaded transcription model; you can download it again later.

During a download, the settings show progress and a cancellation control. If downloading or verification fails, use Try Again on the same screen. Follow the displayed recovery message; Report an Issue is available if the problem persists. If a new download offer appears, review its size before trying again.

Optional speaker model

Speaker detection uses a separate Nemotron 3 model, managed in the Speaker Detection group on the same settings page. It is optional: installing or removing it is independent of the transcription model.

See Detect and Manage Speakers to install it, choose automatic or manual detection, name speakers, and manage local voice memory.