Luca – multifunctional AI assistant for interviews and meetings

Luca is a free, multifunctional assistant powered by artificial intelligence, created to help you during interviews and business meetings. It can listen to your interlocutor, transcribe speech into text, understand the context of the conversation, and instantly suggest thoughtful answers to questions. With Luca, you’ll feel confident in every interview or meeting, receiving real-time AI support.

Flexible Settings and Personalization

  • Multiple models and instructions: Choose from several GPT models by OpenAI for the optimal response style. You can set your own instructions for the AI assistant—for example, add information from your resume to the context so that answers consider your experience and achievements.
  • Response length adjustment: Control the number of tokens (words) returned by the model so that answers are the desired length and detail.
  • Speech recognition models: Luca supports connecting your own Whisper and Voice Activity Detection (VAD) speech recognition models for higher accuracy. By default, it uses the whisper model ggml-large-v3-turbo-q5_0 and the VAD model silero-v5.1.2-ggml, but you can easily replace them with your own.
  • Dialogue scenario: Flexibly configure the assistant’s behavior to your preferences. For example, set a delay (in milliseconds) before the AI assistant starts answering after your interlocutor stops talking—this helps create a natural pause. Set up automatic interruption of the AI’s answer as soon as your interlocutor speaks again, so the assistant never interrupts a real person.
  • Interruption priority: Set priorities for different input sources. By default, the interlocutor’s voice (Speaker) has the lowest priority, your own voice (User Voice) is medium, and messages in the app chat (User Chat) are the highest. This means AI answers won’t be interrupted by the interlocutor’s voice if its priority is lower, but will always stop when you start speaking or enter a manual message.

Context Memory and Conversation History

Luca has advanced context memory. The assistant remembers the entire conversation flow. For example, at the end of an interview, you can ask, “What was my first question?”—and Luca will immediately answer, referring to the beginning of the dialogue. This ensures continuity and allows you to return to any details discussed.

All conversations are automatically saved locally on your device. You can review conversation history at any time, rename a dialogue for convenience, or delete it if it’s no longer needed. Every sent and received message is also under your control: you can delete individual messages or use the “Reply to message” feature. If you press this button next to any message in the history, the AI assistant will generate a response specifically to that question or remark—very convenient for clarifying or diving deeper into a previously discussed topic.

Voice Communication and Audio Integration

Communicate by voice as easily as in a regular conversation. Luca lets you ask questions by voice through a microphone, just as your interlocutor speaks. Your question will be instantly transcribed into text and understood by the AI assistant. Similarly, the assistant’s answers can be voiced (if this option is enabled in the settings), so you get a live interlocutor ready to help in real time.

Flexible audio device setup: You can choose which device will serve as the source of the interlocutor’s voice (Speaker) and which as your own voice (User Voice). On macOS, you can use the virtual audio driver BlackHole to capture sound from the application where the call or conference takes place. For example, just create a new Multi-Output Device in the Audio MIDI Setup utility, add BlackHole as the main device and your speakers (such as MacBook Pro Speakers) as the second. Set the sample rate to 16 kHz and enable Drift Correction for the secondary device—after that, Luca will be able to “hear” your interlocutor directly, while remaining invisible.

The application supports multilingualism—you can specify the language your interlocutor speaks and your own language to improve the accuracy of Whisper speech recognition. Luca easily handles transcription and responses in different languages, making it a universal tool for international teams and interviews.

Invisible Mode

Luca is designed to help you without drawing unnecessary attention. The application works in invisible mode: its window is not shown during screen sharing. If you are asked to share your screen during an interview, your interlocutor will not see the Luca window—it’s only visible to you. On macOS, the assistant can also completely hide from the Dock, running exclusively in the background. All these parameters are easily adjusted in the app settings (press Command + , on Mac or Ctrl + , on Windows to open the settings window).

Privacy and Security

We understand how important privacy is, especially when discussing personal career matters during interviews. Luca does not send your data to third-party servers or save conversation history in the cloud. All information—from audio recordings to text transcriptions and messages—is stored only locally on your computer. You have full control over your data: if necessary, you can always delete the history or individual fragments of the conversation manually.

Advanced Features: Chat and Markdown

In addition to the voice assistant, Luca provides a convenient chat interface. You can write messages directly to the AI assistant, like in a regular messenger, if the voice format is not suitable at the moment. The chat also allows you to attach images (just paste a picture from the clipboard using Command/Ctrl + V) for the assistant to analyze and provide feedback or comments. This opens up new possibilities—from analyzing charts and diagrams to discussing presentation slides right during the meeting.

Luca understands Markdown formatting in messages. This means AI assistant responses can include headings, lists, tables, links, code blocks, and other formatting elements. For example, you can ask for a list or a table-formatted answer and get a beautifully formatted result that’s easy to read. Such Markdown support is especially useful when analyzing saved conversations later or exporting answers for reports.