Introduction
When you add content to your platform, the system analyzes the file to extract textual information. Depending on the format, the platform retrieves this information through text extraction (for documents, images, web files, and similar formats) or transcript generation (for audio, video, and supported learning packages).
Only content from which the system can successfully extract or generate text can be used by platform features that rely on textual analysis:
Content reaches these features from three sources, and each source has its own supported file types and limits:
- Knowledge collections: files synchronized from external sources such as Google Drive and SharePoint, see Knowledge Collections.
- Local files: files you upload from your device, for example in a chat or when configuring a custom agent.
- Learning content: training materials and assets in your platform, including those in courses and learning plans.
This article outlines the supported content types and the conditions required for successful text extraction and transcript generation for each source.
Knowledge collections
By attaching a knowledge collection all of its content becomes available for the agent. For every conversation users have the agent can retrieve information from the attached collection to provide a reply or perform its task.
Content is added to collections from systems connected via Knowledge connectors. Which content type is supported and what content is extracted from the respective type is different for each system. For detailed information, see Knowledge connectors and the detailed documentation for every connector linked there.
Local files
Local files are files you upload directly from your device, when attaching them in a chat with the AI assistant or an agent or adding them as knowledge to a custom agent. Unlike learning content and knowledge collections, local files are not added to your platform. They are used only in the chat or agent where you add them.
The following table lists all file types that can be analyzed by the platform when you upload them as local files, with the maximum size of each file.
| Category | Types | Extracted content | Maximum file size |
|---|---|---|---|
| Document files | .pdf, .pptx, .docx | Text | 50 MB |
| Image files | .png, .jpeg, .jpg, .gif, .webp, .heic | Image | 4 MB |
Learning content
Learning content includes the training materials and assets in your platform, as well as the courses and learning plans that collect them. The following table lists all file types that can be analyzed by the platform when they come from learning content.
| Category | Types | Extracted content | Training materials / Assets |
|---|---|---|---|
| Text files | .txt, .csv | Text | Training materials and assets |
| Document files | .doc, .docx, .odt, .ppt, .pptx, .pdf, .xls, .xlsx | Text | Training materials and assets |
| Image files | .bmp, .jpeg, .png, .tiff | Text in the image | Training materials and assets |
| Web files |
.html, .htm Note: When a web page URL is provided, the transcript is generated only for that specific page. Content from links embedded within the page is not extracted. |
Text | Training materials and assets |
| Audio files | .acc, .mpeg, .wav | Audio transcription | Training materials and assets |
| Video files | .mp4, .mov | Audio transcription | Training materials and assets |
| Google workspace files✴ | Docs, Sheets, Slides | Text | Training materials and assets |
| Linked online videos✴ | YouTube, Vimeo, Wistia | Subtitles | Training materials and assets |
| E-learning packages✴ | SCORM and xAPI/TinCan (Articulate Rise and Articulate Storyline) | Text and audio transcription | Training materials |
| Docebo files | Creator lessons | Text and audio transcription | Training materials |
✴Private content (content requiring authentication to be accessed) is not supported
For the maximum file size of each learning content type, platform-level limits apply, see Supported file types in Docebo system requirements.
Learning content types not listed in the table above are not supported for text extraction or transcript generation. These include assignment, Docebo Learning Impact (DLI), LTI, observation checklist, survey, test, Elucidat, archive, playlist, Shape, AICC, and content from Content marketplace partners.
In addition to being a supported file type, the system must also be able to extract text or generate a transcript from the learning content. If text extraction fails, the content cannot be used by features that rely on textual analysis.
Text extraction or transcript generation may fail in the following cases:
- Audio or video files that contain no speech (for example, background music only)
- Transcripts shorter than 30 words, which are discarded
- Private content that requires authentication to be accessed
- Images compressed to a degree that prevents accurate Optical Character Recognition (OCR).