Recital

What is Recital?

Recital is a platform that helps you capture, classify, and search through documents and emails. Founded in 2017, the team behind it was one of the first international groups to use large pre-trained language models for document processing. These models help computers understand the meaning behind words and phrases, so they need less data to learn. Recital combines text and layout analysis with language models to handle complex documents efficiently.

Who created Recital?

Recital was started in the spring of 2017 by Gilles Moyse, who has a PhD in computer science and serves as CEO, and Frédéric Allary, the General Manager. Gilles is an engineer from UTC, and Frédéric graduated from Sciences-Po and HEC with a degree in econometrics. The company offers a complete document processing solution built on large language models, handling everything from extraction to classification. The team brings together expertise in AI and software development across multiple languages and technical areas, and the company encourages innovation by giving team members time to work on their own technology projects.

What is Recital used for?

• Automate how mail and documents move through your systems.
• Capture the specific information you need from any kind of document.
• Search for answers across your entire collection of documents.

Who is Recital for?

• Professionals who want to automate their document processes.
• Anyone working directly with document processing, including challenging or complex workflows.

Deployment and pricing

Recital can be installed on your own servers (on-premises) or used as a Software as a Service (SaaS) offering. Pricing options are available on the company’s website so you can choose what fits your usage.

Key capabilities

Recital’s strength comes from applying large language models to intelligent document processing. The models’ ability to understand context and generalize means the platform can handle extraction, sorting, and searching across varied document types without needing extensive training data for each one.