Innosign Lab

We try out new tools for Deaf people and the public.

In Innosign Lab we use artificial intelligence to build apps for the Deaf community, research and tools for work that keeps repeating. We are still testing most of them. For each one we say what stage it is at, what automation does by fixed rules, what AI does and what a person decides.

What the status means

Testing
We are trying it out. It is not finished yet.
Research
For now it is research, not a finished tool.
Internal tool
We use it in our own work.
Open source
The code is public. Anyone else can use and adapt it too.

Apps

  • DeafChat

    Status: TestingOpen source planned

    An iPhone app for talking with a hearing person. What the hearing person says appears as captions. What you type, the app reads aloud.

    Who does what

    1. Automation

      Testing. We are preparing it so that only a short summary remains after the conversation.

    2. AI

      It turns speech into captions. If you choose an online service, the audio is sent to its provider.

    3. Person

      You type the reply; the app only reads it out.

  • Furniture AR

    Status: TestingOpen source planned

    An iPhone and iPad app. It shows furniture from a 3D model, for example a kitchen or a wardrobe, in the real room, exactly against a wall or in a corner. It shows its status by colour and vibration; nothing is signalled by sound alone.

    Who does what

    1. Automation

      It snaps the model with its back to the wall and its base to the floor and remembers its place in the room.

    2. AI

      While scanning, it recognises the walls, the floor and the corners of the room. It runs right on the phone.

    3. Person

      You choose the model and the wall it should stand against, and fine-tune the position millimetre by millimetre.

Research and advocacy with the help of AI

  • We advocate that the driving licence of a Deaf person should depend on how safely they drive, not on percentages of hearing loss. Today the decree requires a hearing aid for a hearing loss of 51 to 90% (by Fowler). With a loss above 90%, a driver of a car or a motorcycle (group 1) is assessed individually. A driver of a lorry or a bus (group 2) with a loss above 90% is unfit to drive, unless a special assessment is made for certain hearing implants.

    Source: Decree No. 9/2009 Coll., Annex 5, on Slov-Lex (in Slovak), as of 9 November 2026.

    Who does what

    1. Automation

      It checks that every link to a source in the documents opens.

    2. AI

      It helped to find and compare research and rules in 11 EU countries. While checking, it also found claims that do not hold up. We corrected or removed them.

    3. Person

      We base our claims on verified sources and mark data that is not verified. We prepared the documents with other organisations of Deaf people, and we decide every step.

Tools for our work

For repetitive work we use our own tools. Here are the main ones. Some of them we are only trying out so far.

  • Project finances

    Status: Internal tool

    Budgets of projects and of the company, payments, invoices and receipts, payroll, travel expenses, cash vouchers and contracts in the Central Register of Contracts.

    Who does what

    1. Automation

      It loads the payments from the bank and prepares payment orders, QR codes for payment and cash vouchers. It calculates travel expenses by the Slovak rules.

    2. AI

      It reads invoices and receipts and suggests which project and budget line they belong to. From the e-mails it puts together a list of what needs to be paid.

    3. Person

      We confirm or correct every suggestion. We enter bank payments ourselves in online banking.

  • Project management

    Status: Internal tool

    The status of every project, shared folders with partners, work plans, timesheets, evaluation reports, payment requests and checks of the contracts with partners.

    Who does what

    1. Automation

      For a new project it prepares a shared folder for the partners and a work plan. It fills in timesheets and reports in the required templates.

    2. AI

      It goes through e-mails, documents, the calendar and the project budget and shows what is done and what is missing.

    3. Person

      From the overview we choose the next steps and agree them with the partners.

  • Captions for videos

    Status: Internal tool

    We upload captions for the videos from our projects to YouTube for a whole list of videos at once. Depending on the situation, AI prepares them from the speech in the video or a person writes them.

    Who does what

    1. Automation

      It uploads the finished captions for the videos from our projects to YouTube all at once, not by hand one by one.

    2. AI

      Depending on the situation, AI prepares the captions by recognising the speech in the video, or a person writes them.

    3. Person

      A person decides whether AI or a person prepares the captions, and uploads the finished file. Before uploading we check that every file belongs to the right video.

  • Tracking draft laws

    Status: Testing

    Every day it looks through new draft regulations and alerts us to those that may concern Deaf people.

    Who does what

    1. Automation

      It follows new draft laws and decrees on the Slov-Lex portal and sends us an overview every morning.

    2. AI

      It marks the drafts that may concern Deaf people or accessibility.

    3. Person

      We read the marked drafts and decide what to do about them.

  • Grant calls

    Status: Testing

    It collects new grant calls and helps us choose those that fit our work.

    Who does what

    1. Automation

      Every day it collects new calls from the European Commission portal and from eurofondy.gov.sk. It sorts them by the keywords of our topics.

    2. AI

      For new calls it prepares a short summary, a recommendation and the risks.

    3. Person

      We choose the calls that make sense and decide whether to apply.

  • Innosign Studio

    Status: Testing

    An internal tool for the videos from our projects. From a video it prepares a suggested edit, captions, versions for social media and the text of a post. It saves only a draft to the post scheduler and publishes nothing by itself. We are still testing it.

    Who does what

    1. Automation

      It suggests cuts only in the pauses between captions and prepares versions of the video for each network.

    2. AI

      When needed, it suggests captions from the speech in the video and the text of the post in Slovak or English. By default it runs on our own computer; the cloud only with separate consent.

    3. Person

      We check and edit every version of the video and the text. A person confirms publication and its time.

  • innosign-ui

    Status: Internal toolOpen source planned

    A shared accessible look for our web apps: light and dark mode, high contrast, three text sizes, Slovak and English. Our apps for travel expenses, budgets and grant calls use it.

    Who does what

    1. Automation

      Before every change it checks the contrast of all colours and the accessibility of the screens in every mode.

    2. AI

      No AI is used. The kit calls no model.

    3. Person

      We set the design rules in one document; the kit only applies them. What the automatic check cannot judge, for example buttons with only a symbol, we check ourselves.

  • FluentiaGrant

    Status: Testing

    Experiment: a platform for managing grant projects from the application to the report, with the budget, expenses and tasks. We are trying out whether it suits our work.

    Who does what

    1. Automation

      It shows at any time how much of the project budget has been spent. Rules from past corrections assign new items without AI.

    2. AI

      On import it reads the budget from the application (PDF or Excel) and suggests the budget lines. It can run in the cloud or on your own server.

    3. Person

      We check the AI suggestions and confirm or correct them. The system learns from the corrections.

Open source

We will publish some of our tools as open source. Their code will be public, so other organisations of Deaf people will be able to use and adapt it too.

Open source planned

We plan to publish these first:

  1. DeafChatan app for talking with hearing people
  2. innosign-uian accessible look for web apps

We will publish the code under the MIT licence. A published tool will then carry the label Open source

People lead, AI does the work

Many people think that AI does everything by itself. It does not. We set the goal and the method, and AI works to our plan, not the other way round. We check every result.

We will teach you too. In our training you will try out how to give AI a task, guide it step by step and check what it has done.

There is a fee for the training. We can also prepare it as part of working together in your project. Sometimes it is free, when one of our projects makes it possible. We announce such dates in our News.