Labvanced vs WebGazer
Part of Labvanced's platform comparisons.
Labvanced is a no-code behavioral research platform with its own webcam eye tracking built in. WebGazer.js is a free, open-source JavaScript library from Brown University that estimates where a person is looking from their webcam. Researchers add it to their own code or use it through a platform that integrates it, such as jsPsych or Gorilla. Both run gaze estimation in the participant's browser, but they differ in accuracy, in what they measure beyond raw gaze, and in how much of the study you have to build yourself.
Accuracy is where the two differ most. Labvanced uses its own neural network, which reached an accuracy of 1.4° of visual angle in a peer-reviewed lab comparison with an EyeLink 1000 (Kaduk et al., 2024, Behavior Research Methods). WebGazer's original paper reports an average error of 4.17° in the lab (Papoutsaki et al., 2016), about three times as large. The two figures come from separate studies, so the At a Glance table shows how each was measured.
Beyond gaze, Labvanced detects fixations, lets fixations on areas of interest control the experiment, and can record facial emotions and remote heart rate (rPPG) in the same webcam session. WebGazer returns gaze predictions only. Everything else comes from a host platform or your own code.
Labvanced studies run in the browser, in mobile apps for Android and iOS, and offline in a desktop app for Windows, macOS, and Linux with Lab Streaming Layer (LSL) integration for EEG and other recordings. Its Web Bridge Chrome extension records gaze and mouse activity on external websites.
In short, Labvanced is a maintained platform where gaze measurement, calibration, fixation detection, and the experiment itself already work together, with no code to write. WebGazer is a free, open-source building block that has been validated independently in several journal studies. Its developers stopped official maintenance in February 2026: the library still works, but bug fixes and updates are no longer guaranteed.
Table of Contents
At a Glance
| Capability | WebGazer.js | Labvanced |
|---|---|---|
| No-code visual experiment builder | No: a JavaScript library. No-code use is possible through a platform that integrates it, such as Gorilla | Yes |
| Flexibility to build any paradigm | Yes, but with your own JavaScript or a host platform (jsPsych, Gorilla, PCIbex) | Yes, in the visual editor, with gaze-contingent events and areas of interest (rectangles, polygons, imported SVG shapes) |
| Peer-reviewed evidence | Yes | Yes: Kaduk et al. (2024), Behavior Research Methods |
| Webcam eye tracking accuracy (laboratory) | 4.17° average error against a Tobii EyeX tracker, with a chin rest, from 4 analyzed participants (Papoutsaki et al., 2016); 3.94° offset in a fixation task (Semmelmann & Weigelt, 2018) | 1.4° overall and 1.3° at screen center, from 19 analyzed participants recorded at the same time as an EyeLink 1000 (Kaduk et al., 2024) |
| Webcam eye tracking accuracy (online) | Accurate enough to tell apart the four quadrants of the screen in a visual world study (Slim & Hartsuiker, 2023) | 1.8° median across 494 Prolific participants on their own computers (Eye Tracking 2.0 Lab Note, Labvanced's own validation with open data) |
| Fixation detection | No: the library outputs gaze predictions, and fixations have to be computed separately | Yes: built in, with fixation-based triggers and metrics such as dwell time and time to first fixation |
| Calibration | Learns from the participant's clicks and cursor movements; calibration screens come from the host platform or your own code | A setup wizard with accuracy tiers (from a 30-second to a 4 to 5 minute calibration), a separate mode for infants and toddlers, a virtual chinrest that pauses the study when the head drifts, and gaze control checks during the study |
| Emotion detection | No | Yes: from the same webcam session |
| Remote heart rate (rPPG) | No | Yes: from the same webcam session |
| EEG / LSL hardware integration | No | Yes: via the desktop app |
| Multi-user / multiplayer studies | No | Yes: with lobby management, role assignment, and group formation |
| Desktop app (offline) | No | Yes |
| Mobile app | No | Yes |
| Works on external websites | Only on websites where you can add its code | Yes, on live websites you don't control, through the Web Bridge Chrome extension |
| Support channels | Community support through GitHub Issues | Documentation, live chat, email support, and onboarding calls, depending on the plan |
| Maintenance | As of February 24, 2026, official maintenance has ended. The library still works, but updates are no longer guaranteed | Actively developed; Eye Tracking 2.0 and its online validation were released in October 2026 |
The accuracy rows come from different studies with different hardware, participants, and reference trackers, so they are not a head-to-head test. They are the figures published for each system.
Where WebGazer Is Strong
WebGazer made webcam eye tracking in the browser practical for research. It has been open source since February 2016, and it is free to use under GPLv3. A lab can run it without a license budget, inspect every line of the code, and change it.
It has also been validated independently: groups unconnected to its developers have tested it in peer-reviewed journals.
- Cognitive science: Semmelmann and Weigelt (2018), in Behavior Research Methods, ran fixation, pursuit, and free-viewing tasks in the lab and online, and replicated the well-known pattern of looking most at the eyes when viewing faces.
- Psycholinguistics: Slim and Hartsuiker (2023), in Behavior Research Methods, replicated a visual world study online, and Van der Cruyssen et al. (2024), in the same journal, replicated three established eye-tracking effects with it.
- Developmental research: Steffan et al. (2024), in Infancy, validated it in a multi-lab study of toddlers aged 18 to 27 months.
Its use in published research reflects this. In a scoping review of webcam eye tracking in psycholinguistics, 28 of the 33 experiments that named their software used WebGazer.js (Patterson, Nicklin & Vitta, 2025).
It is also easy to add to an existing setup. jsPsych ships a WebGazer extension with plugins for camera setup, calibration, and validation, so a lab that already runs jsPsych studies can add eye tracking without changing tools. Like Labvanced, WebGazer processes the webcam image in the participant's browser and never sends video to a server.
Where Labvanced Differs
Accuracy. In laboratory testing, WebGazer's published error is about three times larger than Labvanced's: 4.17° (Papoutsaki et al., 2016) against 1.4° (Kaduk et al., 2024). The two figures come from separate studies. WebGazer's was measured on 4 participants against a Tobii EyeX, and Labvanced's on 19 participants recorded at the same time as an EyeLink 1000. Three of the four Kaduk et al. authors are affiliated with Scicovery GmbH, the company that develops Labvanced. Labvanced's online follow-up, Eye Tracking 2.0, reports a median accuracy of 1.8° across 494 Prolific participants on their own computers, and publishes its data and analysis scripts. Accuracy decides how close together your areas of interest can be. Van der Cruyssen et al. (2024) concluded that online WebGazer studies work for relatively large effects that don't need high precision, for example designs with four or fewer large regions of interest.
Measurement, not just gaze coordinates. WebGazer returns a stream of predicted gaze positions, not fixations. Gorilla's documentation for its WebGazer-based eye tracking notes that the estimates "can jump around quite a bit," and says that in its experience "less than 30%" of participants give data good enough to detect fixations, saccades, or blinks (Gorilla Support, eye tracking documentation, accessed 2026-11-13). Slim and Hartsuiker (2023) also measured a delay of roughly 300 ms in WebGazer's eye-movement time course, which matters for questions that need fine timing. Labvanced has built-in fixation detection, and its data export includes fixation counts, fixation durations, dwell time, and time to first fixation per area of interest.
The experiment and the eye tracker are one system. With WebGazer, the experiment logic, calibration screens, areas of interest, data storage, and quality checks come from a host platform or from your own JavaScript. In Labvanced they are part of the same no-code visual editor. A fixation on an area of interest can start the next stimulus through the Eyetracking Fixation Trigger, and areas of interest can be rectangles, polygons, or imported SVG shapes. Calibration is set up through a wizard: you choose who your participants are and how much accuracy you need, and Labvanced sets the calibration length, the virtual chinrest that pauses the study when a participant moves out of position, and gaze control checks that keep the calibration on target. Infants and toddlers get their own calibration mode, with animal images and sounds instead of spoken instructions.
Emotion and heart rate from the same webcam. WebGazer measures gaze only. Labvanced's emotion detection runs from the same webcam session used for eye tracking, with accuracy above 90% averaged across emotions, though accuracy varies between individual emotions. The same session can also capture remote heart rate (rPPG), expected to be within 1 to 3% of actual heart rate. Both are Labvanced's own algorithms, and an independent validation study against gold-standard references is still outstanding. All three measures are processed on the participant's device, so no image or video leaves it.
Group studies. WebGazer has no multi-user support, so a group study needs its own server and synchronization code. Labvanced's multi-user infrastructure adds lobby management, role assignment, and group formation, so several participants can take part in the same study session in real time.
Deployment beyond one web page. WebGazer runs wherever you can add its script to a web page. Labvanced's eye tracking works on desktops, tablets, and phones, and Labvanced studies also run in native Android and iOS apps and offline in a desktop app for Windows, macOS, and Linux. The desktop app connects to EEG and other lab hardware through Lab Streaming Layer (LSL), so webcam eye tracking can be recorded alongside other signals in in-lab sessions. WebGazer was originally built to track gaze on ordinary websites. For website studies, Labvanced's Web Bridge Chrome extension records gaze, clicks, and scrolling on a live external website inside a controlled study session, with areas of interest mapped to page elements.
Support and ongoing development. WebGazer's support is community help through GitHub Issues, and as of February 24, 2026 its official maintenance has ended. The project's README also notes that the current version no longer corresponds to the one described in its publications, so the 2016 accuracy figure describes an earlier version of the library. Labvanced offers documentation, live chat, email support, and onboarding calls, depending on the plan. Its eye tracking is maintained by the team that develops the platform, and its second generation, Eye Tracking 2.0, was released with its online validation in October 2026.
Choose Which Fits Your Study
WebGazer is a good fit if you have no software budget, already run your studies in jsPsych or PCIbex, and your design uses a few large areas of interest, such as a four-quadrant visual world display. It also suits developers who need full access to the source code. Check two things before you start a long project: who will maintain the code now that official maintenance has ended, and whether its spatial accuracy fits your stimuli.
Labvanced is the stronger choice once a study needs more than a rough gaze estimate. It fits when your areas of interest are closer together than about 4°, WebGazer's published lab error, when you need fixations rather than raw gaze samples, or when gaze should drive the experiment itself, for example starting the next trial once a participant fixates a target.
It also fits when eye tracking is one part of a larger design. A single Labvanced study can combine gaze with emotion detection and heart rate from the same webcam, bring several participants together in one session, run on phones and tablets or offline in the desktop app, and record EEG through LSL. On WebGazer, each of these would mean a separate tool or your own code.
Labvanced also saves the time a WebGazer setup costs. Calibration, data storage, quality checks, and fixation analysis are already built in, so a study can be built in the visual editor without writing JavaScript, and Labvanced's team can help through support or the study building service. In a methods section, a Labvanced study can cite both a peer-reviewed lab validation and an open-data online validation.
Migration Note
A study built with WebGazer, for example in jsPsych, cannot be imported into Labvanced directly. It has to be rebuilt in the Labvanced editor. For labs moving existing paradigms, Labvanced's study building service can rebuild the experiments for you. For planning a new design, see How to Run a Webcam Eye Tracking Study and Design Considerations that Affect Webcam Eye Tracking Accuracy.
FAQ
Last Reviewed
Last reviewed: 2026-11-13
WebGazer's license, maintenance status, and technical details on this page were checked against its official website (webgazer.cs.brown.edu), its GitHub repository and source code (github.com/brownhci/WebGazer), and the jsPsych and Gorilla eye-tracking documentation as of this date. Labvanced capabilities were checked against the Labvanced platform codebase and license documentation. Peer-reviewed research citations are referenced inline throughout.