Automation, security and precision – a modern speech-to-text tool

At FireNet we have started work on a new solution that could significantly improve the work of law enforcement, experts and data analysts: an application for the automatic transcription of audio files and voice recordings into editable Word and Excel formats.

This tool is designed for environments isolated from the Internet, including those operating under confidentiality classification. Using artificial intelligence, it will be possible to convert audio material into text documents quickly and securely, without the need for manual transcription.

Project assumptions

The system's goal is the full automation of speech transcription from various audio sources:

audio recordings in popular formats (WAV, MP3, FLAC),
a microphone connected directly to the device,
recordings from interviews, surveillance and field recorders.

The application being developed will make it possible to:

quickly convert speech into an editable document (.docx, .xlsx),
work offline, without Internet access,
install in air-gapped (isolated) environments – in line with IT security requirements,
handle the Polish language with high accuracy, including specialist vocabulary (e.g. legal or technical).

Security first

Given the nature of FireNet's clients – investigative units, public institutions and the administration sector – the whole project has been designed from the outset for full compliance with security policies, such as:

no Internet connection,
no external data processing,
full control over the working environment (systems under confidentiality classification),
the option of local installation on the client's secured hardware.

Current stage: planning and prototype development

The project is currently at an early stage. The team of developers and AI specialists is designing the system architecture, preparing the test environment and evaluating the available language models that will be able to run locally – without the need to send data outside the system.

In parallel, we are working on a user interface that will allow you to:

easily add audio files,
edit and format the transcript,
export data to the chosen format (Word, Excel),
mark key fragments of speech (e.g. in investigations).

What next?

In the coming months we plan to:

create a working prototype for internal testing,
calibrate speech-recognition accuracy in investigative conditions,
extend support to recognising different voices (multi-speaker),
consult institutional partners on operational requirements.

💬 Transcription is a time-consuming process that can consume hours of officers' or experts' work. With FireNet's new tool, this work will be faster, safer and fully automated – with no compromise on data confidentiality.

If your institution is interested in taking part in the testing phase or would like to learn more – please get in touch.