1. Click Local model in the left navigation.
2. Choose a model folder. You can use the same folder as HaoOne if you already have its models.
3. Choose an accessible download source, download a transcription model, and wait for Installed.

Start with Qwen3-ASR-0.6B for a lighter model, or choose Qwen3-ASR-1.7B when recognition quality matters more.

4. Check that the alignment model for your media language is installed.
5. Install WeSpeaker to identify speakers. For semantic search, install both the text embedding model and the reranking model.

Download the emotion and audio-event models if you need those recognition options.