Attico has developed an AI video content analysis tool that automatically converts recordings into structured documentation. The solution combines recording, transcription, context analysis, screenshot selection, and manual generation in a single process, reducing manual effort and accelerating the preparation of materials for users.
Customer
Attico
Industry
IT
Services
AI agents & automation
The solution was originally developed as an internal AI-powered video analysis tool to simplify and reduce the effort required to prepare documentation from video content. To streamline content creation and reduce manual effort, Attico built an AI-powered system capable of transforming video recordings into reusable materials for different audiences, including technical teams, educational institutions, and organizations working with research and medical content.
Teams create valuable knowledge through video recordings, screencasts, lectures, briefings, and product demos. Turning this content into documentation, summaries, and training materials often remains a manual process spread across multiple tools. This may create the following difficulties:
Attico transformed the manual and fragmented process of working with video content into a unified AI pipeline designed for the automatic creation of structured materials. The solution integrates the collection of visual, audio, and browser context, content processing, and the generation of the final document.
Attico developed an AI solution that allows users to convert video into text, transforming the manual creation of documentation from video recordings into a single, structured workflow. This included:
Attico automated the time-consuming process of turning video recordings into structured content. The solution speeds up documentation creation, reduces operational costs, and enables secure content processing based on the client’s requirements. It also can be used to summarize, analyze, and generate various types of content, helping companies streamline diverse workflows.
The solution is built around a Next.js web application, a browser extension, an upload API, and a file-processing worker. Each component handles a different part of the workflow, moving the recording from raw video and browser events to a structured final document.
The browser records video and audio as a WebM file, while the extension captures clicks, form inputs, page transitions, and other UI events. Every event is timestamped from the start of the recording, so it can later be matched to the right moment in the video and transcript.
The processing worker uses FFmpeg to extract the audio track from the WebM file. Whisper then transcribes the audio locally, allowing the original recording to remain within the company’s infrastructure.
The transcript is combined with browser events and other contextual data before being passed to an LLM. The model cleans up the spoken text, organizes it into clear steps, and identifies the points in the recording where screenshots would add useful context. The worker then pulls those frames directly from the video.
The processed text, links, and screenshots are assembled into a user manual. In the current implementation, the system generates Markdown and DOCX files, with DOCX documents created based on a selected template. Intermediate files and processing statuses are stored on disk and are available through a REST API.
Early versions of the solution sometimes generated duplicate screenshots and did not always correctly match transcripts with user actions. To improve the system, the team compared the automatically generated documents with existing documentation and continuously refined the prompts and processing logic. This helped improve transcription accuracy, screenshot selection, and the overall document structure.
Attico provides generative AI development services to help you facilitate data analysis and automate time-consuming communication workflows and content creation.
