The only multilingual speech recognition solution that is fully embeddable (as in “no connection required”)
and 100% GAFAMs-free, with 100% GDPR / CCPA compliance.
Thanks to meaning and unique AI technologies, Total Speech solves major speech recognition pain points:
User side
- Poor level of understanding
- Obligation to adapt / restrain speech (unnatural)
- No *smart* interactions with speech-enabled system (keywords)
- Privacy invasion (GAFAMs)…
Business side
- Limited/impossible embeddability within systems and devices
- Traditional APIs = latency + network required
- Generalist models = poor results on special purpose apps
- GDPR / CCPA compliance…
State-of-the-art Speech Recognition.
In one or several languages
In any UI, application or assistant
100% embeddable in most devices
100%
GAFAMs-free
100% GDPR / CCPA compliant by design.
The ultimate vocal enabler
Highest accuracy (see our industry-standard benchmarks below)
Shines in secure or network-free environments
Easily fine-tunable for special purpose applications (vs generalist competition)
Easily coupled with custom applications based on natural language.
You don’t have to settle for an expensive, uncontrollable GAFAMs-based solution.
Benchmarks – WER* metric
ENGLISH
Models | TotalSpeech | Google ASR (Commercial API) |
|---|---|---|
Librispeech | 3,9% (without LM) 2.96% (with LM) | 12,87% |
CommonVoice | 17,64% | 33,45% |
MLS | 14,5% | 46% |
FRENCH
Models | TotalSpeech | Google ASR (Commercial API) |
|---|---|---|
TV TestSet | 16,6% | 18,21% |
CommonVoice | 17,95% | 35,09% |
MLS | 17,7% | 29,21% |
*WER = Word Error Rate (lower is better).
Get Started Today
Harper Stinson provides deployment-ready AI solutions
together with enterprise-wide transformation programs.