You can detect specific words or phrases spoken and extract high-level information from audio recordings (e.g.: telephone calls, call centers, TV / radio stations or a simple recorded speech or talk).
Information can be used to locate various business issues (e.g.: product issues, marketing issues, finance issues, etc.).
Our products are high-tech software components with complex mathematical algorithms to analyze speech / voice signals.
All algorithms are optimized in order to maximize accuracy at high speed and low resource consumption (CPU and memory).
Our products have no overhead (no database, no complicated setup process), but they focus on technologically complicated tasks.
The runtime modules are just one DLL file with data files leaving a minimal footprint.
The technologies can be integrated via an easy to use Software Development Kit.
We work in a mutual dependency with our integrators as partners, because our partners develop solutions for their market (e.g. call centers)
We align our product roadmap to our partners’ requirements
We focus on core-technology development and not on solution development (marketing)
Compure is not a big multinational company with a big marketing and sales armada, but rather a small team of technicians with high technology skills
We feel to be an ideal extension of our partners, covering complex mathematical know-how to extract information from voice recordings
ACTNow enables to search very fast and accurately for words or phrases in audio or video streams
Create an index and search in it (efficient if you want to do various spontaneous analysis)
Online searches (Search for words or phrases immediately when feeding audio data)
Results contain start time in ms, end time in ms, probability and channel information for stereo input signals
Allows storage of data in any format. The index is a BLOB that can be managed by the integrator (e.g. store in database or simply on hard drive or wherever the integrator wants to)
Support for different resolutions to balance between index size and accuracy
Balance between search speed and accuracy
Different languages: US-English, UK-English, German, French,
Hungarian, Arabic, etc.
Different audio formats
If you have a sound file and a corresponding text (transcript of the sound file), then you can use SynKit to automatically assign time stamps in the audio file for each word in the text
Of course the result depends on the audio quality and the accuracy of the transcript. But even if both are not ideal, SynKit can create very good results
It is up to the integrator where to collect the audio data from. It may come directly from a microphone or a sound file or wherever. The integrator feeds the audio data to SynKit to retrieve the alignment information
If the audio data is in a sound file and the text in a text file, then you can use a ready made function to synchronize both files.
Our partners integrate and redistribute our software components to end-users.
Telephone/Call centers recorders TV/Radio/Broadcast recording Various recordings (meetings, dictation, etc)
We offer different license models to adjust to the model of our partners
We work with a only a few partners to be able to support a very close cooperation to ensure a win/win/win partnership
A big part of our algorithms require a very deep understanding of various mathematics
The analysis of speech signal requires a comprehensive knowledge of signal processing
The linguistic part is important to model each language
An optimal implementation of our algorithms is required to optimize the code for speed, resource consumption and robustness to deliver bug-free software to our partners
Compure Services Kft.
Szépvölgyi út 52, 1025 Budapest,
Hungary, Europe
contact@compure.com
+36 30 991 2487
Tax number: HU22676281