Taking pictures or videos using a third-party app is fairly straightforward,
using ACTION_IMAGE_CAPTURE
or ACTION_VIDEO_CAPTURE
. However, you as the
developer have little control over what happens with the image or video,
other than indicating where the result gets stored. Plus, different camera
apps have slightly different behavior, meaning that you are prone to getting
inconsistent results.
Taking pictures or videos using the built-in Camera
class directly is
eminently possible, but is full of edge and corner cases, not to mention
its own set of per-device idiosyncracies. As a result, a ton of code is
required to successfully show a preview, take a picture, and take a video.
CWAC-Camera
is an effort to standardize that "ton of code" and hide it
behind a scalable API. Here, "scalable" means "simple things are simple,
but complex things may be a bit complex".
This Android library project is also available as a JAR.
If you are upgrading a project using CWAC-Camera to a new edition of the library, please see the "Upgrading" section below.
Step #1: Download the JAR and put it in the libs/
directory of your
project (or, if you prefer, clone this GitHub repo and add
it as a library project to your main project).
Step #2: Add a CameraFragment
to your UI. You have two versions of
CameraFragment
to choose from:
-
com.commonsware.cwac.camera.CameraFragment
for use with native API Level 11+ fragments -
com.commonsware.cwac.camera.acl.CameraFragment
for use with the Android Support package's backport of fragments and ActionBarSherlock, supporting API Level 9 and 10
(note: if you choose the latter, your project will also need to have the ActionBarSherlock library project)
The CameraFragment
is responsible for rendering your preview, so
you need to size and position it as desired.
Step #3: Call takePicture()
on the CameraFragment
when you want
to take a picture, which will be stored in the default digital photos
directory (e.g., DCIM
) on external storage as Photo_yyyyMMdd_HHmmss.jpg
, where
yyyyMMdd_HHmmss
is replaced by the current date and time.
Step #3b: Call startRecording()
and stopRecording()
on the
CameraFragment
to record a video. NOTE that this is presently
only available on com.commonsware.cwac.camera.CameraFragment
for use with native API Level 11+ fragments. The resulting video
will be stored in the default videos directory (e.g., Movies
) on external storage as
Video_yyyyMMdd_HHmmss.mp4
, where
yyyyMMdd_HHmmss
is replaced by the current date and time.
Step #4: Add android:largeHeap="true"
to the <application>
element in the manifest (a requirement which will hopefully be
relaxed in the future).
And that's it.
CameraFragment
(and its underlying CameraView
)
will handle:
-
Showing the preview using an optimal preview frame size, and managing the aspect ratio of the on-screen preview
View
so that your previews do not appear stretched -
Dealing with configuration changes and screen rotation, so your camera activity can work in portrait or landscape
-
Following the appropriate recipes for taking still pictures and videos, including choosing the largest-available image size for the resolution
-
Opening and closing the camera at the appropriate times, so when you are in the foreground you have exclusive camera access, but other apps will have access to the camera while your activity is not in the foreground
-
And more!
Of course, there are probably plenty of things that you will want to configure
about the process of taking photos and videos. There are many hooks in CWAC-Camera
to allow you to do just that.
Much of this configuration involves creating a custom CameraHost
. CameraHost
is your primary interface with the CWAC-Camera
classes for configuring
the behavior of the camera. CameraHost
is an interface, one that you are
welcome to implement in full. Most times, though, you will be better served
extending SimpleCameraHost
, the default implementation of CameraHost
,
so that you can override only those methods where you want behavior different
from the default.
Given a customized CameraHost
implementation, you can pass an instance
of that to setHost()
on your CameraFragment
, to replace the default.
Do this in onCreate()
of a CameraFragment
subclass (or, if practical,
just after instantiating your fragment) to ensure that the
right CameraHost
is used everywhere.
There are a series of methods that you can override on SimpleCameraHost
to control where photos and videos
are stored once taken. These methods will be called for each takePicture()
or startRecording()
call, so you can create customized results for each
distinct photo or video.
Specifically:
-
Override
getPhotoFilename()
to return the base name of the file to use to store the photo -
Override
getPhotoDirectory()
to return the name of the directory in which to store the photo -
Override
getPhotoPath()
to return the completeFile
object pointing to the desired file in the desired directory (the default implementation combines the results ofgetPhotoDirectory()
andgetPhotoFilename()
, so overridinggetPhotoPath()
replaces all of that)
There are equivalent getVideoFilename()
, getVideoDirectory()
, and
getVideoPath()
for controlling the output of the next video to be taken.
By default, if you are using SimpleCameraHost
, your image will be indexed
by the MediaStore
. If you do not want this, override scanSavedImage()
to return false
in your SimpleCameraHost
subclass. This is called on a
per-image basis.
If you override useFrontFacingCamera()
on SimpleCameraHost
to return
true
, the front-facing camera will be used, instead of the default rear-facing
camera.
Or, override getDeviceId()
(available on CameraHost
), and you can provide
the ID of the specific camera you want. This would involve your choosing an
available camera based on your own criteria. See the JavaDocs for Android's
Camera
class, notably
getNumberOfCameras()
and [getCameraInfo()
](http://developer.android.com/reference/android/hardware/Camera.html#getCameraInfo(int, android.hardware.Camera.CameraInfo))
for more.
By default, the pictures taken from the front-facing camera are a mirror
image of what is shown on the preview. If you wish for the front-facing
camera photos to match the preview, override mirrorFFC()
on your CameraHost
and have it
return true
, and CWAC-Camera
will reverse the image for you before
saving it.
There are some exceptions that are thrown by the Camera
class (and kin, like
MediaRecorder
). Those are passed to your host's handleException()
method. The default implementation displays a Toast
and logs the message
to LogCat as an error, but you probably will want to replace that with
something else that integrates better with your UI.
From a UI standpoint, the CameraFragment
solely handles the preview pane.
Presumably, you will need
more to your UI than this, such as buttons to allow users to take pictures or
record videos. You have two major options here:
-
You can put that UI as a peer to the
CameraFragment
, such as by having action bar items, as the demo apps do. -
You can subclass
CameraFragment
and overrideonCreateView()
. Chain to the superclass to get theCameraFragment
's own UI, then wrap that in your own container with additional widgets, and return the combined UI from youronCreateView()
.
You can call autoFocus()
on CameraFragment
or CameraView
to trigger any
auto-focus behavior that you have configured via setFocusMode()
on Camera.Parameters
.
You can call
cancelAutoFocus()
on CameraFragment
or CameraView
to ensure that auto-focus
mode has been canceled.
Note that auto-focus is only available in certain conditions, notably when
the preview mode is enabled. You can call isAutoFocusAvailable()
on CameraFragment
or CameraView
to determine if auto-focus is presently available for use. Calling
autoFocus()
when auto-focus is not available will have no effect.
CameraHost
implementations will need to implement an onAutoFocus()
method, coming
from
the Camera.AutoFocusCallback
interface
that CameraHost
extends.
SimpleCameraHost
has a default implementation of onAutoFocus()
that
plays a
device-standard sound upon completion (API Level 16+ only).
CameraHost
implementations will also need autoFocusAvailable()
and
autoFocusUnavailable()
methods, to be notified when auto-focus is available or
not. This can be used to trigger whether action bar items are enabled, etc.
SimpleCameraHost
has no-op implementations of these callbacks.
By default, the result of taking a picture is to return the CameraFragment
to preview mode, ready to take the next picture. If, instead, you only need
the one picture, or you want to send the user to some other bit of UI first
and do not want preview to start up again right away, override
useSingleShotMode()
in your CameraHost
to return true
.
You will then
probably want to use your own saveImage()
implementation in your
CameraHost
to do whatever you want instead of restarting the preview.
For example, you could start another activity to do something with the
image. However, bear in mind that an Intent
is limited to ~1MB, and so
passing an image to another activity via a Intent
extra is likely to be
unreliable. You will need to do something else, such as (carefully) use a
static data member.
Preview mode will re-enable automatically after an onPause()
/onResume()
cycle of your CameraFragment
, or you can call restartPreview()
on your
CameraFragment
(or CameraView
).
In addition to the configuration hooks specified above, you can do more to tailor how photos and videos are taken.
Your CameraHost
will be called with mayUseForVideo()
, to determine if the preview
should be optimized for possible video recording, or not (i.e., only still images will
be taken).
In the latter case, your CameraHost
will be called with getPreviewSize()
, where you need to return
a valid Camera.Size
indicating the desired size of the preview frames. getPreviewSize()
is passed:
-
the display orientation, in degrees, with 0 indicating landscape, 90 indicating portrait, etc.
-
the available width and height for the preview
-
the
Camera.Parameters
object, from which you can determine the valid preview sizes by callinggetSupportedPreviewSizes()
The CameraUtils
class contains a pair of static methods with stock algorithms for
choosing the preview size:
-
getOptimalPreviewSize()
uses the algorithm found in the SDK camera sample app -
getBestAspectPreviewSize()
finds the preview size that most closely matches the aspect ratio of our available space
SimpleCameraHost
uses getBestAspectPreviewSize()
for the default implementation
of getPreviewSize()
. You can override getPreviewSize()
and substitute in your
own selection algorithm. Just make sure that the returned size is one of the ones
returned by getSupportedPreviewSizes()
.
If mayUseForVideo()
returns true
, though, CameraHost
supplies the preview
size via getPreferredPreviewSizeForVideo()
instead of getPreviewSize()
. If you
wish to use a different preview size for video, return it, otherwise return null
and we will use the results from getPreviewSize()
instead. getPreferredPreviewSizeForVideo()
is passed a Camera.Size
as a hint from the device for a value to use, instead of
anything you might get yourself from Camera.Parameters
-- while using the hinted
value is probably a good idea (if it is not null
), it is not required.
Similarly, your CameraHost
will be called with getPictureSize()
, for you to return
the desired Camera.Size
of the still images taken by the camera. You are simply passed the
Camera.Parameters
, on which you can call getSupportedPictureSizes()
to find out
the possible picture sizes that you can choose from.
The CameraUtils
class has a pair of methods for simple algorithms for choosing a picture
size:
-
getLargestPictureSize()
returns theCamera.Size
that is the largest in area -
getSmallestPictureSize()
returns theCamera.Size
that is the smallest in area
SimpleCameraHost
uses getLargestPictureSize()
for the default implementation
of getPictureSize()
. You can override getPictureSize()
and substitute in your
own selection algorithm. Just make sure that the returned size is one of the ones
returned by getSupportedPictureSizes()
.
When setting up the camera preview, your CameraHost
will be called with
adjustPreviewParameters()
, passing in a Camera.Parameters
. Here, you can make
any desired adjustments to the camera preview, except the preview size (which you
should be handling in getPreviewSize()
). adjustPreviewParameters()
returns
the revised Camera.Parameters
, where the stock implementation in
SimpleCameraHost
just returns the passed-in parameters unmodified.
Shortly after you call takePicture()
on your CameraFragment
,
your CameraHost
will be called with
adjustPictureParameters()
, passing in a Camera.Parameters
. Here, you can make
any desired adjustments to the parameters related to taking photos,
except the image size (which you
should be handling in getPictureSize()
). adjustPictureParameters()
returns
the revised Camera.Parameters
, where the stock implementation in
SimpleCameraHost
just returns the passed-in parameters unmodified.
Shortly after you call startRecording()
, your CameraHost
will be called
with:
-
configureRecorderAudio()
-
configureRecorderProfile()
-
configureRecorderOutput()
in that order. Here, you can help tailor the way videos get recorded.
Each of these is passed the ID of the camera being used for recording plus
the MediaRecorder
instance that does the actual recording.
The stock SimpleCameraHost
does the following:
-
In
configureRecorderAudio()
,SimpleCameraHost
callssetAudioSource(MediaRecorder.AudioSource.CAMCORDER)
on theMediaRecorder
-
In
configureRecorderProfile()
,SimpleCameraHost
callssetProfile(CamcorderProfile.get(cameraId, CamcorderProfile.QUALITY_HIGH))
on theMediaRecorder
-
In
configureRecorderOutput()
,SimpleCameraHost
callssetOutputFile(getVideoPath().getAbsolutePath())
on theMediaRecorder
(wheregetVideoPath()
was described earlier in this document)
While these are reasonable defaults, you are welcome to override these implementations to do something else.
The default SimpleCameraHost
logic for saving photos uses the getPhotoPath()
and related methods discussed above. Actually saving the photo is done in
saveImage(byte[])
, called on your CameraHost
, where SimpleCameraHost
has a
saveImage(byte[])
implementation that writes the supplied byte[]
out to the desired
location.
You are welcome to override saveImage(byte[])
and do something else with the byte[]
,
such as send it over the Internet. saveImage(byte[])
is called on a background thread,
so you do not have to do your own asynchronous work.
Another use for this is to find out when the saving is complete, so that you can
use the resulting image. Just override saveImage(byte[])
, chain to the superclass
implementation, and when that returns, the image is ready for use.
There is also a saveImage(Bitmap)
callback, giving you a decoded Bitmap
instead of a byte[]
. To use this, there is a second version of takePicture()
that you can call that takes two boolean
parameters, indicating whether or
not you want the saveImage(Bitmap)
callback called and/or the
saveImage(byte[])
callback called. The zero-argument takePicture()
indicates
that you only want saveImage(byte[])
called. If you pass true
as the
first parameter to the two-parameter takePicture()
method, then your host
will be called with saveImage(Bitmap)
. Note that if you do this, you are
responsible for the Bitmap
(e.g., calling recycle()
on it) once it is handed
to your host.
Your CameraHost
implementation can return a Camera.ShutterCallback
object
via getShutterCallback()
,
which will be used in the underlying takePicture()
call on the Android Camera
,
giving you control to play a "shutter click" sound. SimpleCameraHost
returns null
from getShutterCallback()
, to give you the device default behavior.
Device cameras are generally set up to take landscape pictures. If you try to use a camera to take a picture in portrait mode, one of three things will happen:
-
Everything works fine, with the device capturing a portrait image
-
The device crashes (this scenario CWAC-Camera aims to handle)
-
The device captures a landscape image, but sets an EXIF header to indicate that an image viewer should rotate the image to portrait when displaying it
The problem with the last scenario is that not all image viewers will honor this EXIF header.
Your CameraHost
will need to implement rotateBasedOnExif()
to indicate
if you want the library to rotate the image automatically. If you return
true
, you will get a portrait image from all devices, for all three
of the above scenarios. If you return false
, the last scenario will
be ignored, and you may get a portrait or a landscape image.
SimpleCameraHost
returns true
for rotateBasedOnExif()
.
CameraHost
exists to provide a hook for you to determine how your app
should handle taking pictures and videos. DeviceProfile
, on the other hand,
provides information about how the device handles taking pictures and videos.
Different devices do slightly different things when working with the camera.
Sometimes this is based on API level, sometimes it is based on how the device
manufacturer tinkered with Android, and sometimes it is based on the underlying
camera hardware. DeviceProfile
provides a place for the CWAC-Camera project
to isolate these differences.
CameraHost
has a getDeviceProfile()
method that should return an instance
of the DeviceProfile
to use for the device that is running the app.
The implementation of getDeviceProfile()
on SimpleCameraHost
calls the static getInstance()
method on DeviceProfile
, which chooses
a DeviceProfile
based on internal heuristics. If you encounter problems
with certain devices, you can detect those in your getDeviceProfile()
method
and return a DeviceProfile
that addresses your needs, otherwise settling for
using the library's own choice of DeviceProfile
.
At present, there are four methods on DeviceProfile
that you can tailor
in your subclasses:
-
useTextureView()
should returntrue
ifCameraView
should use aTextureView
for rendering the preview frames, orfalse
if aSurfaceView
should be used instead -
encodesRotationToExif()
indicates if the device puts information about the device orientation into EXIF headers of the JPEG image -
rotateBasedOnExif()
should returntrue
if the library should attempt to physically change the orientation of the image if the EXIF orientation header indicates that the image should be changed,false
otherwise -
getMaxPictureHeight()
returns the maximum image height to be selected byCameraUtils.getLargestPictureSize()
, to work around devices that report invalid largeCamera.Size
values
If you wish to eschew fragments, you are welcome to work with CameraView
directly. To do this:
-
Add it in Java code by calling its one-parameter constructor, taking your
Activity
as a parameter. At the present time,CameraView
does not support being placed in a layout resource. -
Call
setHost()
on theCameraView
as early as possible, to make sure that theCameraView
is working with the rightCameraHost
implementation. Alternatively, overridegetHost()
and return the rightCameraHost
there. -
Forward the
onResume()
andonPause()
lifecycle events from your activity or fragment to theCameraView
.
Otherwise, CameraView
should work as a regular View
.
These are above and beyond the bugs filed for this project:
-
Taking videos in portrait mode will result in the video files still being stored as landscape, but with a bit in the MPEG-4 header indicating that the output should be rotated. Unfortunately, many video players ignore this header. This is a function of how
MediaRecorder
works, and there is no current workaround inCWAC-Camera
for this behavior. -
Taking photos in portrait mode, for some devices, will have a similar effect: the photo is saved in landscape, with an EXIF field in the JPEG indicating that the results should be rotated.
CWAC-Camera
detects this and tries to correct it, so the image is saved in portrait. However, this may consume too much memory at present, which is why Step #4 above calls for you to addandroid:largeHeap="true"
. This will hopefully be rectified in a future version of this component. -
While a picture or video is being taken, on some devices, the aspect ratio of the preview gets messed up. The aspect ratio is corrected by
CWAC-Camera
once the picture or video is completed, but more work is needed to try to prevent this in the first place, or at least mask it a bit better for photos. -
The Samsung Galaxy Ace refuses to honor a portrait preview in an activity that itself supports portrait or landscape. If you lock your activity to only display in landscape, the Galaxy Ace will probably work.
If you are moving from an older to a newer edition of CWAC-Camera, here are some upgrade notes which may help.
CameraHost
now requires implementers supply mayUseForVideo()
(true
if the
preview should be optimized for possible use in video recording) and
getPreferredPreviewSizeForVideo()
(returns the preview size to use in case
mayUseForVideo()
returns true
). SimpleCameraHost
provides stock implementations
of these, but if you created your own CameraHost
from scratch, you will need
to add your own versions of these methods.
CameraHost
now extends Camera.AutoFocusCallback
, requiring an implementation
of onAutoFocus()
. SimpleCameraHost
shows a basic implementation that, on
API Level 16+, plays the device-standard "hey! you're focused now!" sound.
Developers moving from v0.0.x to v0.1.x should note that you now need to pass
a Context
into the constructor of SimpleCameraHost
. This can be any Context
,
as SimpleCameraHost
retrieves the Application
singleton from it, so you do not
have to worry about memory leaks.
- Acer Iconia Tab A700
- Amazon Kindle Fire HD
- ASUS Transformer Infinity (1st generation)
- Galaxy Nexus
- HTC Droid Incredible 2
- HTC One S
- Lenovo ThinkPad Tablet
- Nexus 4
- Nexus 7 (1st generation, 2012)
- Nexus 7 (2nd generation, 2013)
- Nexus 10
- Nexus One
- Nexus S
- Motorola RAZR i
- Samsung Galaxy Note 2
- Samsung Galaxy S3
- Samsung Galaxy S4 (GT-I9500)
- Samsung Galaxy Tab 2
- SONY Ericsson Xperia Play
- SONY Xperia E
- Sony Xperia S LT26i
- SONY Xperia Z
This project depends on the Android Support package and ActionBarSherlock
at compile time, if you are using
the Android library project. It also depends on the Android Support package and
ActionBarSherlock at runtime
if you are using the .acl
flavor of CameraFragment
.
This is version v0.4.0 of this module, meaning it is rather new.
In the demo/
sub-project you will find a sample project demonstrating the use
of CameraFragment
for the native API Level 11 implementation of fragments. The
demo-v9/
sub-project has a similar sample for the CameraFragment
that works
with ActionBarSherlock.
The code in this project is licensed under the Apache Software License 2.0, per the terms of the included LICENSE file.
If you have questions regarding the use of this code, please post a question
on StackOverflow tagged with commonsware
and android
. Be sure to indicate
what CWAC module you are having issues with, and be sure to include source code
and stack traces if you are encountering crashes.
If you have encountered what is clearly a bug, or if you have a feature request, please post an issue. Be certain to include complete steps for reproducing the issue.
Do not ask for help via Twitter.
Also, if you plan on hacking on the code with an eye for contributing something back, please open an issue that we can use for discussing implementation details. Just lobbing a pull request over the fence may work, but it may not.
- v0.4.0: fixed bug in
getBestAspectPreviewSize()
, added hooks for device overrides for video preview sizes, improved support for HTC One - v0.3.0: improved support for auto-focus, Samsung Galaxy Camera, etc.
- v0.2.1: CyanogenMod devices will now use
SurfaceView
regardless of API level - v0.2.0: auto-focus support, single-shot mode, Droid Incredible 2 fixes
- v0.1.1: improved support for Nexus 4 and Galaxy Tab 2
- v0.1.0: Nexus S crash fixed, added support for indexing images to
MediaStore
- v0.0.4: Nexus S EXIF issue fixed, added
saveImage(Bitmap)
callback - v0.0.3: shutter callback support, bug fixes
- v0.0.2: bug fixes
- v0.0.1: initial release