Frequently Asked Questions
Which use cases does natif.ai offer?
Can I test the technology beforehand?
You can test the various models free of charge on our platform .
There you have a monthly free quota of approx. 100 pages. However, depending on the workflow you use, the quota may vary.
How much does the software cost?
Can natif.ai be used on-premise?
Under certain conditions, our technology can also be used on-premise. If you are interested in a server-based model, please feel free to make an appointment.
Where can I look at customer feedback?
If you want to get an idea of our past performance, feel free to check out our customer-feedback.
What is Deep OCR?
Deep-OCR is our innovative deep-learning based technology for Optical Character Recognition (OCR). Compared to existing algorithms, our Deep-OCR can recognize text much better even if the document is low quality (crooked, wrinkled, shadows, etc.).
How does the natif.ai AI work?
Our AI uses textual and visual information to classify documents and recognize key information, just like a human.
Unlike most solutions on the market, our AI is based on neural networks that continuously learn and develop a general understanding of documents. Therefore, our models can be used directly without complex configuration, while the numerous rule-based approaches of the competition usually have to be configured laboriously and expensively to your own data first.
You can use our pre-trained models or pure OCR, and you can also train your own workflow to classify your documents.
How exactly does the AI read the documents?
Our AI models have been trained with hundreds of thousands of documents and are thus able to read out all important information and position data with uniquely high precision on the market. Thanks to Active Learning, the AI continuously learns from inaccuracies and errors and thus becomes more and more precise and better.
Can OCR also read handwriting?
Yes, our Deep OCR technology is capable of recognizing handwriting in addition to machine-generated documents, so it can be used for a wide variety of work steps and workflows. For example, you can easily automate the reading of hand-filled forms and receipts with our AI.
Does the OCR also recognize individual line-item data?
Yes, our technology is capable of reading a wide variety of position data. In addition to classic information from the header and footer, our AI can also extract position data with uniquely high accuracy on the market.
How long does the software take to read a document?
Our RESTful API processes all documents in real-time. Since our AI runs on GPUs, we process documents parallel in the shortest possible time (usually less than 1 second per page).
What languages does the technology support?
Our OCR supports German and English as well as various European languages. The selection of languages depends on the respective extraction model.
Basically, our text recognition can read all characters of the Latin alphabet with high accuracy. On request we can extend our models to other languages.
Can you process Excel lists?
Yes, if they have been transferred to an image file.
Can you process Emails?
Yes, if they have been transferred to an image file.
What output formats do you have?
ZugFerd, Json, XML on request, PDF with OCR Layer
Can you read QR and barcodes?
Yes we can, however, this is not available in our standard plans.
How much data is needed to train a new document type?
This depends on the document type and the number of data fields. More complex documents (invoice, expert opinion, etc.) need up to 1000 documents, simple cases like forms only a few (5-10).
Can I train my own AI model?
Yes. On our platform you can train your own model for example your own document extraction.
How many documents are needed to train an AI model?
The amount of data depends on the complexity of the respective use case, the diversity of documents and the individual customer needs. In cooperation with our experts, our technology is optimally integrated into existing workflows and systems. Please feel free to make an appointment for further information.
Do the documents require a certain quality?
No. Our AI has been trained to recognize and read documents of varying quality – just as a human would. So it can read even scans of crumpled documents or handwritten documents with extreme precision.
Do I have to preprocess my documents in a certain way?
Unlike older OCR systems, our solution does not require documents to be pre-processed. We free them from the background for you, correct alignment errors, and read with the highest precision despite low image quality.
What are the minimum requirements for the documents to be processed ?
A4 format min. 220 DPI, beyond that the document must be readable for a human being as well.
How many pages may a document have?
We have a limit of about 1,000 pages per document.
Does the document have to be in A4 format?
Where and how long is the data stored?
You can decide for yourself how long the data is stored in our data center in Saarwellingen. Our technology only needs your information for the period of processing – by default, we delete it after two weeks.
Is natif.ai GDPR compliant?
Yes, data protection is very important to us. You can find all information about data protection here.
Where are your servers located?
Our servers are located in two data centers (Saarland and Rheinland-Pfalz).
What is the security level of your data center?
Our data center has level 3.
Who is your data privacy officer?
Mauß Datenschutz from Hamburg.
Where is the data processed?
Exclusively in the EU – in Germany.
Who is the API designed for?
Our API is designed for developers. Deep OCR can be easily integrated as a tool into your existing system and significantly improve the automation rate.
Where can I find the RESTful API documentation?
All important information about the API integration can be found here.
Where can I view the documentation for each workflow?
The documentation for each of our workflows can be found here.
To which CMS/DMS can the API be connected?
In principle, our API can be connected to all common CMS/DMS. The prerequisite for this is the ability to send a request to our API. For example, we offer a Make-Plugin.
How does the integration to other systems work?
The integration depends on the respective use case and the individual customer needs. In cooperation with our experts, our technology is optimally integrated into existing workflows and systems. Please feel free to make an appointment for further information.
Is it possible to use self-created workflows in Make?
Yes, you can use our Make-Plugin for that.