Video Coding Software for Behavioral Researchers
Behavioral video projects can need manual annotation, automated estimates, or both. Choose tools around a written coding task rather than assuming every facial-analysis product answers the same question.
Published and updated November 1, 2026.
Published tool overview as of November 1, 2026
| Tool | Type | License or cost as published | Platforms | Notes | Source |
|---|
| BORIS | Event logging | Free, open source, GPL v3 | Linux, Windows, macOS | Video, audio, live observation; version 10.0.2 released October 16, 2026 | BORIS software |
| ELAN | Time-aligned annotation | See the vendor's site | See the vendor's site | Multiple hierarchical tiers; EAF XML | ELAN |
| Datavyu | Video coding and visualization | Free, open source; check specific terms | Windows and Mac; Java-based | Ruby scripting; builds on OpenSHAPA | Datavyu |
| The Observer XT | Behavior coding and analysis | See the vendor's site | See the vendor's site | Live or recorded video; multimodal sync; Cohen's kappa | The Observer XT |
| FaceReader | Automated facial expression analysis | Essential EUR 2,000; Advanced EUR 9,000; Premium EUR 12,500, excl. tax, as of November 1, 2026 | Desktop software; operating systems not specified here | Webcam or video; editions add AUs and Baby FaceReader | FaceReader |
| OpenFace | Facial-analysis toolkit | Commercial users directed to separate CMU licensing | See the vendor's site | Landmarks, head pose, AUs, eye gaze | OpenFace toolkit and licensing |
| Py-Feat | Python facial-analysis toolbox | MIT code; pretrained weights have separate terms | Python; operating systems not specified here | AUs, emotions, landmarks; version-specific additional outputs | Py-Feat toolbox and documentation |
| Mental Edge | Source-linked expression review API | See developer pricing; no prices stated here | Developer API for recorded-image and video workflows | Asynchronous jobs; time-coded scores for contextual review | Mental Edge developer pricing |
BORIS, as of November 1, 2026
ELAN, as of November 1, 2026
Datavyu, as of November 1, 2026
The Observer XT, as of November 1, 2026
FaceReader, as of November 1, 2026
OpenFace, as of November 1, 2026
Py-Feat, as of November 1, 2026
Mental Edge developer pricing
Separate annotation from automatic measurement
An annotation tool records a reviewer's coding decisions about a source recording. An automated facial-analysis tool estimates features or categories from the material it processes. A project can use both, but their outputs should remain distinguishable.
Write down whether your question concerns events, time intervals, movement intensity, or expression categories. Define the source references and uncertainty labels your reviewers will need. These decisions make the comparison more useful than a general list of features.
The overview reports only published facts from the linked sources, as of November 1, 2026. Missing platform or licensing details are left unspecified rather than filled with assumptions. Confirm those details with each vendor before selecting software or planning a purchase.
BORIS for event logging
As of November 1, 2026, BORIS describes free, open-source event logging for video, audio, and live observation. Its published license is GPL v3, with Linux, Windows, and macOS listed. Version 10.0.2 was released on October 16, 2026.
Friard and Gamba's 2016 paper is the scholarly reference for BORIS. Cite that work when describing the tool in a methods section. Keep the software version and your own coding scheme in the project record as separate information.
Consider an event-logging workflow when reviewers need to mark defined behaviors over time. Test a small example using your intended event definitions before expanding the collection. A well-organized log still depends on clear instructions and dependable coding decisions.
ELAN for layered, time-aligned annotation
As of November 1, 2026, The Language Archive describes ELAN as an audio and video annotation tool. Its published description includes unlimited annotations on multiple hierarchical tiers. Annotations are time-aligned and saved as XML in the EAF format.
That structure is relevant when a project needs separate annotation layers tied to the same recording. Decide which layers belong in your codebook rather than adding categories because the software permits them. Document relationships among layers so another reviewer can understand the annotation design.
This guide does not list platform or license details for ELAN. Check the linked ELAN page for current terms before selecting software. Inspect a sample export to confirm that source intervals and tier relationships survive your planned workflow.
Datavyu for video coding and visualization
As of November 1, 2026, Datavyu describes a free, open-source, Java-based tool for Windows and Mac. Its published description covers video coding and data visualization. It builds on OpenSHAPA, supports Ruby scripting, and lists support from NSF and NICHD.
Consider whether a scripted coding workflow would help your team repeat its own processing steps. Do not assume scripting automatically improves the reliability of movement labels. Keep coding decisions inspectable even when later processing is automated.
The Observer XT for structured behavioral review
As of November 1, 2026, Noldus describes The Observer XT as event logging for live or video behavior. It lists coding schemes, multimodal synchronization, and inter-rater reliability using Cohen's kappa. These are published capabilities, not a claim about the quality of a particular study.
An agreement feature is useful only when the coding unit and comparison rules are clearly defined. Decide which reviewers, clips, and annotations are being compared. Preserve initial independent labels before producing a reconciled version.
This guide does not list a price or operating systems for The Observer XT. Use the linked product page to clarify procurement and platform questions. Avoid borrowing those details from a different Noldus product or an undated comparison.
FaceReader for automated facial-expression estimates
As of November 1, 2026, Noldus describes FaceReader as desktop analysis from webcam or video. Its editions add capabilities including Action Units and Baby FaceReader. The latter is described for ages six to twenty-four months and is based on Baby FACS.
As of November 1, 2026, Noldus lists Essential at EUR 2,000, Advanced at EUR 9,000, and Premium at EUR 12,500. Those published prices exclude tax and describe FaceReader, not Mental Edge plans. Confirm the current edition and terms directly before relying on an old quotation.
This guide does not repeat the vendor's accuracy claim or offer an independent performance assessment. Define how your reviewers will inspect estimates beside the source recording. Keep automated categories separate from participant reports and manual FACS annotations.
OpenFace for facial-analysis components
As of November 1, 2026, OpenFace lists landmark detection, head pose, AU recognition, and eye gaze. Its README directs commercial users to separate licensing through CMU. Check those terms before making assumptions based on repository access.
Use the published capabilities to identify components worth evaluating for your specific task. Do not infer that a component toolkit supplies a finished research method or complete annotation workflow. Plan how source intervals, quality exclusions, and independent labels will be maintained around it.
Py-Feat for Python-based exploration
As of November 1, 2026, Py-Feat describes AU, emotion, and landmark detection from images and video. Its version-one description uses sixty-eight-point landmarks. Version two adds valence and arousal, gaze, and a 478-point MediaPipe mesh.
As of November 1, 2026, its code is MIT-licensed, while pretrained weights have separate terms. Treat those licensing questions separately when planning research reuse or a commercial implementation. Do not assume a code license automatically covers every model asset.
Choose outputs that match the question and document any mappings between labels. An emotion category, a landmark position, and a movement code are not interchangeable observations. Keep the source recording available for checking unexpected outputs and annotation disagreements.
Where Mental Edge fits
Mental Edge surfaces observable expression and movement signals from recorded video for contextual human review. Its asynchronous API returns time-coded scores that reviewers can inspect beside source moments. Those patterns may indicate intervals for review, not proof of feelings or truthfulness.
The developer API also includes face detection, embeddings, and a face gallery. Published facial-expression job states are Queued, Processing, Completed, and Failed. See developer pricing for commercial details rather than treating this guide as a price comparison.
Mental Edge does not replace an annotation codebook, independent review, or a legal assessment. It is not for hiring, promotion, diagnosis, deception detection, threat assessment, student scoring, or surveillance. Select it only where its defined review role fits an appropriate purpose.
Choose a workflow before choosing a tool
Compare source navigation, annotation structure, exports, licensing, and the review team's actual needs. Try representative clips, including difficult visibility conditions, before processing an entire collection. Keep manual labels and software estimates separate during that evaluation.
FACS guidance recommends agreement between at least two independent coders. Barrett and colleagues caution that facial configurations do not uniquely establish internal emotion. A useful workflow therefore preserves source context, uncertainty, and the distinction between observation and interpretation.