Introducing Guardian Pulse 2.0

Guardian Pulse 2.0 is OQENYX's first native multimodal model — and it ships today. Audio, video, and text are processed as first-class inputs, unified in a single endpoint with Guardian's privacy-first architecture.

Here is what Pulse 2.0 can do right now: transcribe and analyse audio up to 10 hours per request in 113 languages, process video up to 400 seconds at 720p with full scene understanding, generate natural speech in 36 languages, and handle joint audio-visual tasks like meeting summarisation, captioning, and multilingual voice Q&A. It handles speech recognition in 113 languages, text-to-speech in 36 languages with strong voice quality, and a broad range of audio-visual tasks.

Now for the honest part. Pulse 2.0 launches with certain capabilities available via the Onora App and a subset of features not yet enabled at the consumer level. Specifically, voice cloning and speaker personalisation — features that allow the model to replicate a specific person's voice from a short audio sample — are implemented and technically ready. We are not enabling them yet.

The reason is straightforward: voice cloning technology carries meaningful consent and misuse risk. We are working through the regulatory and legal framework to do this right — verified consent flows, audit trails, and explicit opt-in requirements that satisfy both GDPR and the emerging AI Act provisions around biometric data. We expect to enable voice personalisation features in a limited early access programme in the coming months, beginning with enterprise customers who have specific verified use cases.

We also have speaker diarisation (identifying who said what in a multi-speaker recording) and real-time streaming audio in active development. These will arrive as incremental updates to the same Pulse 2.0 endpoint — no breaking changes.

What we are not doing is shipping these features half-finished to hit a launch checklist. Pulse 2.0 today is genuinely strong on what it does. The capabilities that are not yet available will ship when we are confident they are safe to ship.

If you have a use case that requires voice personalisation or diarisation, reach out directly. We are prioritising enterprise onboarding for early access to the restricted feature set.