Editorial illustration for Ollama's New Phonon Feature Reads AI Replies Aloud
Ollama Adds Text-to-Speech with Phonon Model
Ollama's New Phonon Feature Reads AI Replies Aloud
Ollama added text-to-speech on October 24, 2026, starting with Phonon, a small text-to-speech model built by Gradium. The feature works right now on macOS, with Windows and Linux versions still in the pipeline. Typing `ollama run phonon "Hello from Ollama."` is enough to get a spoken reply, and users can swap voices with a `--voice` flag pulled from Gradium's voice catalog, viewable in full with `ollama show phonon`.
The bigger draw is what happens when Phonon gets paired with a chat model. Piping a model's text output straight into Phonon lets Ollama read answers aloud as they're generated, so playback starts after the first sentence instead of waiting for the whole response. Output can also be redirected to a WAV file instead of played live, and the whole thing runs through an OpenAI-compatible `/v1/audio/speech` endpoint, so code already built for that API should work with little rewriting.
Licensing varies by model size. Phonon itself ships under Apache 2.0, but the smaller 40M variants are CC BY-NC 4.0, meaning they're restricted to noncommercial projects. Ollama also claims a speed edge on Apple hardware worth a closer look.
Ollama can now turn text into speech. The first speech model is Phonon, a family of small, state-of-the-art text-to-speech models from Gradium. Speech models are now available for macOS, with Windows and Linux support coming soon.
Why this matters
For developers building voice interfaces, Phonon's arrival inside Ollama is a shortcut worth testing immediately. Piping a chat model's output straight into a speech model, with playback starting after the first sentence rather than waiting for the full response, solves a real latency problem that's plagued local voice assistants. That's a concrete engineering win, not a marketing claim.
But macOS-only support, with Windows and Linux "coming soon," means teams building cross-platform products can't rely on this yet. The MLX acceleration numbers on M5 Max hardware are specific to Apple Silicon, so anyone on other chips should treat the throughput claims as not yet applicable to their setup.
For founders, the bigger signal is Ollama's trajectory: it's moving from a model runner into a fuller pipeline tool, chaining text generation and speech synthesis in one command. That's a meaningful shift in scope. Worth watching whether Gradium or others ship comparably sized speech models for other runtimes, and whether Ollama's Linux rollout arrives on a real timeline rather than staying vague.
Related Reading
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
Open Source Box CEO Aaron Levie Says AI Will Drive "Biggest Cybersecurity Upgrade
Open Source Study: 6.7% of Deepfake Requests on Hugging Face Targeted Children
From Archives Cognition Releases Open-Source Agent to Patch Critical Node.js RCE Vulnerability
From Archives Anthropic Says Unreleased Model Outperforms Mythos 5.1 in Capability, Alignment
Common Questions Answered
What is Phonon and when did Ollama add text-to-speech support?
Phonon is a small text-to-speech model built by Gradium that Ollama integrated on October 24, 2026. It enables users to convert text into spoken replies by running simple commands like `ollama run phonon "Hello from Ollama."` and represents Ollama's first text-to-speech feature.
How can users customize voices when using Ollama's Phonon feature?
Users can swap voices using the `--voice` flag, which pulls from Gradium's voice catalog. The full catalog of available voices can be viewed by running the command `ollama show phonon`.
What platforms currently support Ollama's Phonon text-to-speech model?
Phonon is currently available on macOS, with Windows and Linux versions still in development and coming soon. This platform limitation means cross-platform teams building voice interfaces will need to wait for broader support.
What specific latency problem does pairing Phonon with a chat model solve?
When Phonon is paired with a chat model, playback can start after the first sentence rather than waiting for the full response to complete. This solves a real latency problem that has historically plagued local voice assistants, providing a concrete engineering improvement for developers building voice interfaces.
