HapticGen

Tool Summary

Metadata
Release Yearⓘ The year a tool was first publicly released or discussed in an academic paper.	2025
Platformⓘ The OS or software framework needed to run the tool.	Python
Availabilityⓘ If the tool can be obtained by the public.	Available
Licenseⓘ Tye type of license applied to the tool.	Open Source (MIT)
Venueⓘ The venue(s) for publications.	ACM CHI
Intended Use Caseⓘ The primary purposes for which the tool was developed.	Prototyping

Hardware Information
Categoryⓘ The general types of haptic output devices controlled by the tool.	Vibrotactile
Abstractionⓘ How broad the type of hardware support is for a tool. Bespoke: Hardware is custom built, including open hardware. Consumer: Support is limited to a specific commercial device or a family of devices. Class: Any device with the same affordances can be used, possibly with some configuration required.	Consumer
Device Namesⓘ The hardware supported by the tool. This may be incomplete.	Meta Quest
Device Templateⓘ Whether support can be easily extended to new types of devices.	No
Body Positionⓘ Parts of the body where stimuli are felt, if the tool explicitly shows this.	N/A

Interaction Information
Driving Featureⓘ If haptic content is controlled over time, by other actions, or both.	Time
Effect Localizationⓘ How the desired location of stimuli is mapped to the device. Device-centric: No spatial information is taken into acccount. Actuators are directly controlled. Target-centric: Stimuli are placed on the body or are assigned to a virtual object to be triggered through interaction. The tool handles how the device will display the effect. Location-aware: The desired location of an effect is taken into account to some extent, but is ultimately presented in context of the output device.	Device-centric
Non-Haptic Mediaⓘ Support for non-haptic media in the workspace, even if just to aid in manual synchronization.	None
Iterative Playbackⓘ If haptic effects can be played back from the tool to aid in the design process.	Yes
Design Approachesⓘ Broadly, the methods available to create a desired effect. Direct: low-level parameters are directly modifiable. Procedure: parameters are controllable by an abstract process. Additive: reusable effects are ordered in time to create complex effects. Library: a library of pre-existing effects is available for use or re-use. Description: a natural language description of the experience is used to find an appropriate effect, often through searching a library.	Procedural, Description
UI Metaphorsⓘ Common UI metaphors that define how a user interacts with a tool. Track: a timeline represents an interactive channel containing effects. Keyframe: key points of the effect are set and behavior between them is interpolated. Music Notation: an adaptation of a musical score or notation represents haptic effects. Dataflow: a dataflow programming model is used to control haptic output. Demonstration: physical actions or other data are mapped simply to effects or parameters.	None
Storageⓘ How data is stored for import/export or internally to the software.	WAV
Connectivityⓘ How the tool can be extended to support new data, devices, and software.	None

Additional Information

HapticGen uses a modified version of Meta’s AudioCraft project to generate haptic waveforms from user-specified prompts. The AudioCraft model was fine-tuned using a haptic dataset created automatically from an existing text-to-audio dataset that was then curated by expert participants. In the design interface, the duration of a desired effect can be specified by the user, but other aspects of it are left entirely to the model.

For more information, consult the CHI’25 paper and the GitHub repository.