CereProc
Edinburgh voice lab known for regional accents, character voices and voice reconstruction work, with an offline SDK for embedded use
Quick Overview
| Company | CereProc Ltd |
|---|---|
| Category | AI Voice & Speech |
| Headquarters | Edinburgh, United Kingdom |
| Founded | 2005 |
| EU Presence | Yes - United Kingdom (UK GDPR) |
| Data Location | United Kingdom; offline SDK available |
| Open Source | No |
| GDPR Compliant | Yes (UK GDPR) |
| Self-Hosting | Yes - offline SDK for embedded deployment |
| Main Features | Regional accent voices, custom voice creation, voice reconstruction, offline SDK, CereWave AI neural voices, SSML control |
| Pricing | From about GBP 10/month (personal) / custom pricing for commercial and embedded |
| Best For | Projects needing authentic regional accents or reconstructed personal voices |
| Replaces | ElevenLabs, Amazon Polly, Nuance |
Detailed Review
CereProc built its reputation on something the large providers largely ignore: accents that actually sound like the places they come from. Scottish, Irish, Welsh, northern English and a range of other regional voices are core catalogue rather than curiosities, which matters for broadcasters, public bodies and games that would otherwise be stuck with generic received pronunciation.
The company is based in Edinburgh and has also done notable voice reconstruction work, rebuilding voices for people who had lost theirs — including the widely reported reconstruction of Roger Ebert's voice from archive recordings.
Accents and Character
The catalogue includes voices that are recognisably local rather than nationally averaged. For a Scottish public sector announcement system, a regional broadcaster or a game that wants its characters to sound like somewhere specific, this is the practical differentiator over larger vendors.
CereWave AI, the neural engine, brought the naturalness up to modern standards while retaining that accent range, and SSML support gives fine control over pronunciation, emphasis and pacing.
Voice Reconstruction
CereProc has specialised in building voices from limited or imperfect archive material — recordings never made for the purpose of voice cloning. That is technically much harder than cloning from clean studio audio, and it is what makes reconstruction of a deceased or ill person's voice possible at all.
As with any cloning work, consent and rights are the governing constraint, and commercial reconstruction requires clearance from the voice owner or their estate.
Offline Deployment
The SDK runs locally on Windows, macOS and Linux with no network dependency, which suits assistive software, kiosks and any application where confidentiality or availability rules out a cloud call.
Limitations to Consider
CereProc is a small specialist rather than a scale platform: the voice catalogue is narrower than the big cloud providers' and the tooling around it is more basic. Being UK-based, it falls under UK GDPR rather than EU GDPR, which for some EU organisations means an extra transfer assessment. And for pure expressive performance, generative competitors have moved ahead.
Who Should Use CereProc
Broadcasters, public bodies, game studios and assistive software developers who need a voice that sounds like a specific place, or who need a voice reconstructed from archive material. For generic corporate narration, larger providers are cheaper and simpler.
Pros and Cons
Pros
- Genuine regional accents, not generic ones
- Voice reconstruction from archive material
- Offline SDK with no network dependency
- SSML control over pronunciation
- Affordable personal tier
Cons
- Small catalogue compared to cloud giants
- UK rather than EU jurisdiction
- Less expressive than generative rivals
Alternatives to CereProc
Other European ai voice & speech worth comparing before you decide:
Synthesia
United Kingdom · AI video avatars with synthetic voice in 140+ languages
Respeecher
Ukraine · Speech-to-speech voice cloning used in film and television
ReadSpeaker
Netherlands · Enterprise text-to-speech in more than 50 languages
Acapela Group
Belgium · Custom synthetic voices, including assistive personal voices
See all European ai voice & speech or compare them directly against ElevenLabs on our alternatives page.
Frequently Asked Questions
Accent authenticity. CereProc maintains genuinely regional voices — Scottish, Irish, Welsh, northern English and others — where large providers offer nationally averaged accents. If sounding local matters, this is the specialist.
Yes, voice reconstruction from limited archive material is one of its specialities, and it is technically harder than cloning from clean studio audio. Commercial use requires consent from the voice owner or their estate.
Yes. The SDK runs locally on Windows, macOS and Linux without a network connection, which suits assistive software and kiosk applications.
It falls under UK GDPR rather than EU GDPR. The protections are closely equivalent, but EU organisations may need to document the transfer, which is not required for an EU-established vendor.