svara-global-v1 is a 780M-parameter voice model designed to fit how each team actually works. The same voice that carries an ad spot through 14 languages also handles your phone agent, your audiobook narrator, and your game NPCs. One model. One API. One voice. Many industries.
The same voice your audience has heard for 18 months — now in Spanish, Japanese, Portuguese, and 47 more. Same brand. Same script. New languages, in one API call.
[lang=es] mid-script switches language without re-prompting.Cast a single narrator and let them carry your title from cover to cover, in every language your distributor will take. Studio-grade voice, no booking, no re-recording for re-edits.
Author once in your authoring tool. Pick a narrator. The same voice teaches the same course in every language your students speak — with built-in pronunciation hints for technical terms.
Plants take in carbon dioxide through tiny pores called stomata. Inside the leaf, chlorophyll absorbs sunlight and powers the conversion to glucose.
Glucose travels through phloem to nourish the rest of the plant, while oxygen exits through the same pores.
Stream replies under 200ms first-byte; interrupt cleanly when the customer talks; switch language mid-call. Plug into Twilio, Vonage, Daily — keep the same voice.
Cast a handful of voices, tag your dialog with emotion and language, ship in 12 markets without a re-record. Inline [whisper], [shout], [laugh] as part of your script.
Brand-matched voice. 14 languages on day one. Containment up, escalation down — because callers actually understand what they're being asked, in the language they speak.
svara-global-v1 is small enough to self-host on a single A10, expressive enough for audiobook narration, fast enough for live phone agents, and broad enough to serve every market your team works in.
5,000 free characters every month. Open weights, Apache 2.0. From an audiobook chapter to a live phone agent — one model, one voice, one bill.
Start free →