Speecon Suite
From your documents to your server.
The Speecon Suite is our pipeline and toolset for adapting speech recognition to your field. It finds your terms, measures what your system misses, and delivers a reviewed, signed vocabulary pack — or a fine-tuned adapter — that runs on your own hardware. Your experts decide at every checkpoint.
Animated illustration of a German dental dictation appearing word by word, with specialist terms recognised.
Illustrative preview, not a measurement. In a Suite session we run it on your terms and your audio.
One tool for every step.
Customer portal
One personal link for everything your team does with us: open tasks, reviews, report and pack download.
Term discovery
We find the specialist words in your documents and merge them with curated sources from your field. Your own words come first.
Term review
Your experts keep, drop or edit each term. The pipeline waits until you sign off.
Vocabulary audit & report
Each term in several sentences, spoken by several voices, checked on more than one recognition system. A German report with a clear recommendation.
Pronunciation lexicon
Latin, English and abbreviations are proposed the way your field really says them. An expert listens to every proposal and decides.
Vocabulary pack
The reviewed terms become a signed profile that steers recognition towards your vocabulary, without retraining.
Training data
Consent-based recordings with your team, natural sentences spoken by many licensed voices, and audio that sounds like your rooms and noise.
Fine-tuning with a safety net
A small adapter on a proven base model. A forgetting guard refuses any adapter that harms everyday speech.
Serving & lineage
A recognition server on your own GPU, a signed before/after acceptance check, and a record of what was built from what.
Three ways to start.
01
Diagnose
Intake, term discovery, term review and the vocabulary audit with its German report. Answers: does my system know my words?
02
Adapt
Everything in Diagnose, plus the pronunciation lexicon, a signed vocabulary pack, the on-premises server and a signed acceptance check.
03
Fine-tune
Everything in Adapt, plus recording sessions with your speakers, multi-voice and acoustic training data, and a custom adapter with a forgetting guard.