Worth deciding early, because it shapes your whole interaction design rather than being a setting you toggle.
Controllers give you precision, physical buttons, and haptics. You know exactly when a grab starts and ends. They work in poor lighting. Almost every VR game uses them, so players already know how they work.
Hand tracking removes a barrier. No controller to find, no batteries, no explaining what the grip button is. For a demo at an event, a training experience, or anything where someone puts the headset on once, that matters a lot.
What hand tracking is still weak at: precision at speed, anything needing a firm confirmation, occlusion when one hand passes behind the other, and poor lighting. Fast, accurate actions are still better on controllers.
The design consequence people miss: hand tracking has no buttons, so every action needs a gesture or a dwell, and gestures need teaching. Pinch to select is the closest thing to a standard. If you invent your own gesture vocabulary, budget time for onboarding, because nobody arrives knowing it.
A practical middle path: build interactions against an abstraction rather than against a specific input, so a grab is a grab regardless of source. Then support both, and let the headset decide. This costs a little up front and saves a rewrite later.
If you only support one, pick based on who puts the headset on. Repeat users who will learn your game: controllers. First-timers who get one attempt: hand tracking.
Which is your audience?
Controllers give you precision, physical buttons, and haptics. You know exactly when a grab starts and ends. They work in poor lighting. Almost every VR game uses them, so players already know how they work.
Hand tracking removes a barrier. No controller to find, no batteries, no explaining what the grip button is. For a demo at an event, a training experience, or anything where someone puts the headset on once, that matters a lot.
What hand tracking is still weak at: precision at speed, anything needing a firm confirmation, occlusion when one hand passes behind the other, and poor lighting. Fast, accurate actions are still better on controllers.
The design consequence people miss: hand tracking has no buttons, so every action needs a gesture or a dwell, and gestures need teaching. Pinch to select is the closest thing to a standard. If you invent your own gesture vocabulary, budget time for onboarding, because nobody arrives knowing it.
A practical middle path: build interactions against an abstraction rather than against a specific input, so a grab is a grab regardless of source. Then support both, and let the headset decide. This costs a little up front and saves a rewrite later.
If you only support one, pick based on who puts the headset on. Repeat users who will learn your game: controllers. First-timers who get one attempt: hand tracking.
Which is your audience?