News / Models
Work begins on PICO-1, our custom text-to-speech model. A single voice, English-only, tuned for CPU first, and built around a streaming buffer that begins playback the moment the first audio frame is ready.
August 13, 2026
At Sential Innovations, our work has always been about shipping the best models and integrations we can. To that end, we have started development on a new series: PICO. The first model in the series is PICO-1, our custom text-to-speech (TTS) model.
PICO is built to be expressive and aware of what it is producing. Where most TTS systems are judged on output quality alone, PICO is also judged on how quickly it can begin. We treat time-to-first-audio as a first-class deliverable, not a side effect of generation speed.
PICO is designed around a custom streaming system. As PICO generates audio, it produces it in short pieces we call frames. Those frames pass into a buffer that holds them in order until the previous frame has been played. As soon as the buffer has its first frame, playback begins. PICO keeps producing; the buffer keeps handing frames to the audio output in sequence.
The buffer is also a safety layer. Every frame passes through it and is stored there until it is its turn to play, which means frames cannot be lost or outrun the speaker. If PICO slows down under load, frames simply queue in the buffer instead of starving the output. The result is steady audio with no lag spikes, even when the model behind it is under pressure.
PICO-1 is a single-voice model. It currently supports English only; PICO itself cannot speak other languages. Other models you run through Kova support their own languages, but PICO does not, in this release.
PICO-1 will be available via Kova and will run locally. Our current targets for time-to-first-audio (TTFA) are:
These are ranges, not guarantees. Actual rates will vary with the hardware you have available, the length and complexity of the input, and the rest of the workload on the machine. PICO is tuned for CPU first and GPU second, the opposite of how most of our other models are tuned, because the kind of device that benefits most from a local TTS is often the kind of device that does not have a discrete GPU.
Development is underway. The near-term plan is to finish the PICO-1 training and integration work, validate TTFA on a range of hardware, and ship the model through Kova once we are confident in both the output quality and the streaming behaviour. We will share more when there is something concrete to report, not before.
