Google tests granular Gemini voice controls
2026-07-17
Personalization is no longer a perk; it is the product. A hidden settings pane in Gemini’s assistant interface now points to voice controls that break the old “pick a character and live with it” model, exposing sliders for speed, energy, and two additional vocal parameters that shape how the assistant actually sounds in use.
The striking shift is toward signal control, not personality branding. Speed tuning adjusts temporal cadence and prosody, while the energy slider appears to modulate amplitude dynamics and emphasis, sitting on top of the existing text‑to‑speech synthesis stack and its acoustic modeling pipeline. Early screenshots suggest at least four distinct knobs, likely covering speaking rate, loudness contour, pitch, and pause density, moving Gemini closer to a proper prosody editor than a static voice picker.
What looks cosmetic is really strategic infrastructure. By letting users fine‑tune latency‑sensitive traits like rate and pause behavior, Google can better align Gemini with different hardware, accessibility needs, and regional listening habits without maintaining a bloated catalog of prebuilt personas. After recently adding two new default voices, the company now seems ready to treat those voices as starting templates rather than endpoints, hinting at a future in which the assistant’s sound is negotiated by the user, not dictated by the brand.
Loading...
