Thoughts for v2

[natario1]

I would say that v1.4.X is a very stable release, and there's just minor refinements to do.
Major changes would require API signature changes as well so I am starting thinking about a v2 release based on feedback and requests. This is how I see it.

API	Action
Audio enum	Improve it, not just ON and OFF but rather audio options to be passed to the video recorders.
onPictureTaken(byte[])	Refactor to onPictureTaken(PictureResult)
onVideoTaken(File)	Refactor to onVideoTaken(VideoResult)
PictureResult	Add. Candidate APIs: isSnapshot(), getTimestamp(), getLocation(), getOrientation(), getSize(), getJpeg(), toBitmap(maxWidth, maxHeight, callback).
VideoResult	Add. Candidate APIs: isSnapshot(), getTimestamp(), getLocation(), getOrientation(), getSize(), getFile().
setJpegQuality	Remove. It makes no sense, we get a jpeg already from the camera, we can't choose the quality. It is only used when recompressing after post-processing.
setCropOutput	Remove. See below.
getPreviewSize	Change: it should return the size correctly flipped in the view dimensions, and cropped if the preview is cropped (we currently return the camera1 preview size but it's useless.). We should also adapt it to the video size when the session type is video.
getSnapshotSize	Remove. It is equal to getPreviewSize
getVideoSize	Add. Return the size of video outputs taken with takeVideo.
capturePicture	Rename to takePicture. Never crop the output, use the internal picture sizes.
captureSnapshot	Rename to takePictureSnapshot. Takes picture from preview frames. Always crop the output to the visible part. Cropping is already native in YuvImage, but rotating is not. We need either a native jpeg rotator, or (better) a library that writes EXIF orientation tags to jpeg buffers.
startCapturingVideo	Rename to takeVideo.
stopCapturingVideo	Rename to stopVideo.
takeVideoSnapshot	Add. This should be the video equivalent of takePictureSnapshot. It should take videos by collecting preview frames, and crop them to the visible part. So if preview is 1:1, we get 1:1 videos. Then mix with audio. This will be API18 min, see the grafika example.
take* APIs in short:	We would have takePicture, takeVideo, takePictureSnapshot and takeVideoSnapshot. The snapshot version will implicitly crop the output to the visible part, and will be lower quality of course.
VideoRecorder	Create. Since we must record videos in two ways, create an abstract VideoRecorder class. One implementation will contain MediaRecorder stuff that is already working well. The other should contain MediaCodec stuff to create videos from frames. Of course, takeVideo uses the first while takeVideoSnapshot uses the second.
Gesture.DOUBLE_TAP	Add. This is not impossible, add it to the set of supported gestures.

Feel free to discuss. No hurry, I expect this in the middle of 2018.

[ychescale9]

Any timeline for supporting Camera2? Thanks.

[natario1]

I see no urgency for that, so it will come after v2 unless someone works on it.

[tomislavr]

I am using this library for media capture in a application which has encrypted media gallery and encrypts picture on saving it to the file. It would be perfect if possible to get video "stream" directly from your library so we can encrypt it before it gets saved in the file. Is it possible to have takeVideo(OutputStream) call, in any format?

[ghost]

Hey, I noticed that you are adding the label v2-fixed.

For my usecase, I have to set the orientation hint to a specific value for video, but there seems to be no public API for that. Therefore, I currently have to maintain my own fork.

Will v2 contain more public APIs to set the values?
By the way great work and library. Really love this library. :)

[natario1]

Thanks! No, I think we want to keep getting the orientation hint from the device and offer no API.
I don't know why you want to set it yourself but your use case is very rare.