[PTZ] Should pan, tilt, and zoom double constraints apply to getUserMedia?

[beaufortfrancois]

When would web developers set pan, tilt, and zoom double values as part of a getUserMedia call?

await navigator.mediaDevices.getUserMedia({
  video: {
    pan: 1,
    tilt: 2,
    zoom: 3,
  }
});

As we're not sure which camera would be plugged in and cameras have different ranges for pan, tilt, and zoom, I'm not sure those values should be "set" or "applied" to the video stream, as they're not meaningful.
In other words, what does even mean providing pan:1 without knowing the min and max values?

In this case, shall we simply set them as boolean constraint for getUserMedia (or have as suggested initially one single panTiltZoom boolean constraint) and keep double constraint for applyConstraints?

One more thing. Could this code be a thing eventually?
Would it be a way to detect a certain type of camera?

await navigator.mediaDevices.getUserMedia({
  video: {
    pan: { min: 0, max: 1 },
    tilt: { min: 0, max: 1 },
    zoom: { min: 0, max: 1 },
  }
});

[beaufortfrancois]

@reillyeon @eehakkin @riju @guidou

[eehakkin]

When would web developers set pan, tilt, and zoom double values as part of a getUserMedia call?
[...]
As we're not sure which camera would be plugged in and cameras have different ranges for pan, tilt, and zoom, I'm not sure those values should be "set" or "applied" to the video stream.

Web developers could have a priori knowledge of ranges (for instance if a user wants to resume a previously stopped video stream) in which case the site could issue

await navigator.mediaDevices.getUserMedia({
  video: {
    deviceId: "...",
    pan: 1,
    tilt: 2,
    zoom: 3,
  }
});

Even without a priori knowledge, it might make sense to use {video: {pan: 0, tilt: 0, zoom: 1}}, for instance. If the user does not have a PTZ camera, pan/tilt/zoom constraints make no harm and if the user does have a PTZ camera, the constraints most probably reset the PTZ camera to a sensible state. For pan and tilt, the ranges is probably symmetric so 0 is in the middle. For zoom, the value 1 probably means no digital zoom or is less than or equal to the minimum of an optical zoom range resulting in widest possible optical zoom to be selected by media stream selection algorithms.

Resetting a PTZ camera to a sensible state might be quite important because the image quality of a PTZ camera may degrade drastically if a digital zoom level is high.

[beaufortfrancois]

Web developers could have a priori knowledge of ranges (for instance if a user wants to resume a previously stopped video stream) in which case the site could issue

await navigator.mediaDevices.getUserMedia({
  video: {
    deviceId: "...",
    pan: 1,
    tilt: 2,
    zoom: 3,
  }
});

The PTZ camera device could have been placed somewhere else or pointed in another direction, so those values are not guaranteed to still be good.
I thought adding the boolean was to not alter PTZ values. If a video stream is stopped, we can safely assume that when a user wants to resume a previously stopped video stream, it will used its previous position, so no need to manually restore those values.

Even without a priori knowledge, it might make sense to use {video: {pan: 0, tilt: 0, zoom: 1}}, for instance. If the user does not have a PTZ camera, pan/tilt/zoom constraints make no harm and if the user does have a PTZ camera, the constraints most probably reset the PTZ camera to a sensible state. For pan and tilt, the ranges is probably symmetric so 0 is in the middle. For zoom, the value 1 probably means no digital zoom or is less than or equal to the minimum of an optical zoom range resulting in widest possible optical zoom to be selected by media stream selection algorithms.

Resetting a PTZ camera to a sensible state might be quite important because the image quality of a PTZ camera may degrade drastically if a digital zoom level is high.

It seems to me that this is simply a shortcut of this code below.

const stream = await navigator.mediaDevices.getUserMedia({
  video: { pan: true, tilt: true, zoom: true, deviceId: '...' }
});
const videoTrack = videoStream.getVideoTracks()[0];
await videoTrack.applyConstraints({
  advanced: [{ pan: 0, tilt: 0, zoom: 1 }],
});

I think providing a "reset" button to users would be a nice way for websites to reset the camera to a sensible state if that's their wish instead of assuming that this is the case.

Note that on Android, the zoom property resets to "1" each time the camera is "opened"

One thing I've learnt while reading about PTZ is that we can't make guesses about the values of pan and tilt and their meanings. They are even sometimes off.

[reillyeon]

Calling getUserMedia({ pan: true, tilt: true, zoom: true }) and then adjusting based on the reported capabilities seems like the most sensible site behavior but I don't see an issue with sites using double constraints immediately. Is this consistent with other parameters to getUserMedia()?

[beaufortfrancois]

I believe it is how it works with other parameters of getUserMedia but like I said, it seems odd to allow this kind of initial behaviour for pan, tilt, and zoom.

[reillyeon]

I don't see any security or privacy concerns with allowing it even though it doesn't seem that useful.

[beaufortfrancois]

I guess it will be a matter of documenting best practices then. I'll close this non-issue now.
Thanks everyone for your feedback.

[fippo]

If a video stream is stopped, we can safely assume that when a user wants to resume a previously stopped video stream, it will used its previous position, so no need to manually restore those values.

Unless and until Chrome turns off the camera light when track.enabled is set to false, GUM will be frequently called in the same session to emulate Firefox' behaviour (see w3c/mediacapture-main#642)