Deng Plus
Highest fidelity
The full-fidelity engine for published work: narration, audiobooks, and anything a listener plays more than once.
- Rendered 24 kHz WAV exports
- Full-length scripts, chapter by chapter
- Provenance sealed on every file
Deng Flash is live — streaming speech on every plan
Open the studioRead the blogRega Voice
Rega Voice turns a script into speech in Sorani, Iraqi Arabic, Arabic, Turkish, English, Spanish, French, or Chinese — streamed as it renders, sealed with provenance, yours to publish.
Two engines
Same library, same provenance, same credits — pick the engine that fits the job.
Highest fidelity
The full-fidelity engine for published work: narration, audiobooks, and anything a listener plays more than once.
Streams live
The streaming engine: first audio arrives while the rest of the line is still rendering. Built for previews and live calls.
Use it for
Intros, sponsor reads, and full episodes with a cast that stays consistent across a series.
Explainers and ads in eight languages — regenerate one line, not the session.
Long scripts rendered chapter by chapter, provenance intact across the whole book.
The same voices answer live calls through Rega Agent, streamed through Deng Flash.
Turn written Kurdish and Arabic into speech for audiences that listen, not read.
Creator and developer
Script editor, voice rail, streaming previews, and the claim ledger — the whole workflow with nothing to wire up.
Open the AppOne POST renders speech from your code. Same voices, same 24 kHz WAV, same provenance — addressed per render.
Read the API docsQuestions
Sorani Kurdish, Iraqi Arabic, Arabic, Turkish, English, Spanish, French, and Chinese. Kurdish and Arabic are tuned first.
Fidelity versus latency. Deng Plus renders full-fidelity takes for published work; Deng Flash streams first audio immediately and powers previews and live calls.
The demos on this page and in the Labs voice galaxy play without an account. Rendering your own scripts starts with free App credits.
Yes, after a verified verbal consent record signed in the App. One clone per account, and the consent stays attached to the voice.
Every export carries a C2PA manifest and a ledger claim — verifiable without an account, without trusting us.