Talk to Your AI: Voice Mode in InnerZero

You can talk to InnerZero. Not in a gimmicky "say a command and wait" way. Actually talk, like you would to a person. Ask questions, give instructions, have a back-and-forth conversation. Zero listens, thinks, and responds with a natural voice.

And all of it happens locally on your PC.

How it works

Click the microphone icon in InnerZero and start talking. Zero uses local speech recognition (faster-whisper) to transcribe your words, then a fast local AI model to generate a response, then local text-to-speech (Kokoro) to speak the answer.

The whole pipeline runs on your hardware. No audio is sent to the cloud. No third party hears what you say. Your voice data is processed in real time and discarded after transcription.

What you can do with voice

Voice mode isn't just for simple questions. Zero has the same capabilities in voice mode as in text chat:

Hands-free questions. "What's the weather in Manchester?" "How many ounces in a kilogram?" "When is the next bank holiday?" Zero uses its built-in tools to get real answers, not just AI guesses.

Quick tasks. "Set a timer for 15 minutes." "Remind me to call the dentist at 3." "What time is it in Tokyo?" Practical stuff you'd normally reach for your phone to do.

Conversations. Ask follow-up questions. Zero remembers the context of your voice conversation. "Tell me about the Mars rovers." "Which one is still active?" "How long has it been operating?" It tracks the thread naturally.

Dictation. Click the dictation button and speak. Your words appear as text that you can paste anywhere. Useful for writing emails, notes, or messages without typing.

The voice experience

Zero's default voice is powered by Kokoro, a lightweight text-to-speech model. It sounds natural and clear. There are 14 voice options to choose from in Settings.

The speech recognition model (faster-whisper) is remarkably accurate, even with accents. It runs on your GPU if available, giving near-instant transcription.

Response time depends on your hardware. On a machine with a decent GPU, the full cycle (listen, think, speak) takes 2 to 4 seconds. It feels like talking to someone who pauses briefly before answering.

Cloud voice (optional)

If you want even more natural-sounding voices, there's an optional cloud voice mode powered by OpenAI's audio API. This sends your text to OpenAI for voice synthesis, giving you access to 13 ChatGPT voices.

Cloud voice is completely optional. It's disabled by default. It requires a cloud API key. And it only sends the text response to be spoken, not your voice input or conversation history.

Most people won't need it. The local voice is good enough for everyday use.

Privacy first

This is the part that matters. When you talk to InnerZero:

• Your voice is processed locally. No audio recordings are sent anywhere.
• Speech recognition runs on your hardware. No third-party transcription service.
• The AI response is generated locally. No cloud inference.
• Text-to-speech runs locally. No external voice synthesis (unless you opt into cloud voice).

Compare this to Alexa, Siri, or Google Assistant, all of which send your voice to cloud servers for processing. InnerZero keeps everything on your machine.

Getting started with voice

Open InnerZero and look for the microphone icon in the bottom-left of the chat area. Click it to start voice mode. The first time you use it, it loads the speech models (takes about 10 seconds). After that, it's instant.

You might want to adjust the silence threshold in Settings to match how quickly you pause between sentences. The default works well for most people.

For the full picture of what InnerZero can do, read our launch announcement. To see how it compares to cloud alternatives, check out our ChatGPT comparison.