Free AI Tool Now Lets You Talk to Your Computer
Soon you can chat with your devices using just your voice.
Deep Dive
The mtmd component now supports a Gemma 4 input audio stream, according to the changelog. A related server change enables realtime embeddings only for audio generation: the realtime slot previously forced embeddings on, which marked every row of a batch as an output.
Because only a model that generates audio reserves n_batch outputs, a chat model with an audio stream like Gemma 4 aborted on its first batch of audio rows. The hidden state, per the article, is only read by the audio generation pipeline.
Key Points
- llama.cpp, a free AI tool, now supports audio input for the Gemma 4 model.
- This means you can talk to your computer and it will understand you, without typing.
- The update fixes a bug that stopped chat models from handling audio, making voice interaction smoother.
Why It Matters
This brings us closer to hands-free AI, making technology more accessible and natural to use.