Table of Contents
It is possible to stream audio and video to a device via the stream interface using AirPlay. The AirPlay suite consists of two protocols:
- AirTunes/RAOP - Used for real time streaming of audio
- “AirPlay - Everything else (video, images and screen mirroring)
Currently there is some AirPlay functionality supported in pyatv, but it is very limited. These features are currently supported:
- Device authentication (“pairing”)
- Playing media via URL
- Streaming of local files
Early support for streaming audio files via RAOP is also supported (even for non-Apple TV devices). MP3, wav, FLAC and ogg files are supported. Devices that require a password are only supported for the AirTunes/RAOP protocol, not the AirPlay protocol.
In the external interface, AirPlay (including RAOP) support is implemented via the interface.Stream interface.
Using the streaming API
Devices supporting the AirPlay protocol (e.g. Apple TV) can play files by simply providing a URL. It will then be streamed directly from the device. Audio can be streamed to other devices (like AirPlay speakers) that does not support this.
Play from URL
Playing a URL is as simple as passing the URL to interface.Stream.play_url:
url = "http://commondatastorage.googleapis.com/gtv-videos-bucket/sample/BigBuckBunny.mp4" await atv.stream.play_url(url)
If the device requires device authentication, credentials must be present for the AirPlay service. Otherwise an error message will be shown on the screen.
To play a local file, just pass a local file to interface.Stream.play_url instead:
url = "/home/user/BigBuckBunny.mp4" await atv.stream.play_url(url)
When doing this, pyatv will internally start web server on a random port, serving this file only and start streaming from there. When streaming is done, the web server is shut down.
The Apple TV will not provide any feedback if anything is not working. If you have
problems, start by testing the example file above (
BigBuckBunny.mp4) as that is
known to work. Also make sure that you don’t stream from an HTTPS server with a bad
or self-signed certificate: that will not work.
Stream a file
To stream a file, use interface.Stream.stream_file:
stream = ... await stream.stream_file("sample.mp3")
Files in MP3, WAV, FLAC and OGG format are supported and will be automatically converted to a format the receiving device supports. Metadata is also extracted from files of these types and sent to the receiver.
It is also possible to stream directly from a buffer. In this example, a file is read into a buffer and streamed:
import io with io.open("myfile.mp3", "rb") as source_file: await stream.stream_file(source_file)
Streaming directly from
stdin also works, e.g. when piping output from another
stdin is a text stream, the underlying binary buffer must be retrieved and used.
It is also possible to use an asyncio StreamReader as input. Here is an example piping output from ffmpeg:
import asyncio.subprocess as asp process = await asp.create_subprocess_exec( "ffmpeg", "-i", "file.mp3", "-f", "mp3", "-", stdin=None, stdout=asp.PIPE, stderr=None, ) await self.atv.stream.stream_file(process.stdout)
When streaming from a buffer, it’s important to know that some audio formats are
not suitable for that. MP3 works fine, WAV and OGG does not. The reason is that
seeking is done in the stream and
stdin does for instance not support that. If
the buffer supports seeking, then all formats will work fine, otherwise stick with
MP3. For the same reason, metadata will not work if seeking is not supported as
that is extracted prior to playing the file, so seeking is needed to return to
the beginning of file again before playback.
Note 1: Since pyatv v0.13.0, buffer improvements have been made to support some seeking, even in non-seekable streams. This is however not fool-proof and not all audio formats (or files/streams) work, but compatibility is much better from that version and onwards.
Note 2: that there’s (roughly) a two second delay until audio starts to play. This is part of the buffering mechanism and not much pyatv can do anything about.
By default, pyatv will try to extract metadata from whatever content you are playing. Some file formats or streams either does not support nor provide any metadata, in which case you can manually provide the metadata that pyatv will report to the receiver by passing an instance of interface.MediaMetadata when starting to stream:
from pyatv.interface import MediaMetadata metadata = MediaMetadata(artist="pyatv", title="Look at me, I'm streaming") await stream.stream_file("myfile.mp3", metadata=metadata)
All fields in interface.MediaMetadata can be overridden (including artwork) except for interface.MediaMetadata.duration, which is ignored. Please note that artwork must be in JPEG format.
Custom metadata will override any metadata provided by the streamed content. You
can however tell pyatv to only override metadata fields that are missing by setting
from pyatv.interface import MediaMetadata metadata = MediaMetadata(artist="pyatv") await stream.stream_file("myfile.mp3", metadata=metadata, override_missing_metadata=True)
This will use all the metadata from
myfile.mp3, but use pyatv as artist but only
if that field is not present in the file.
Stream from HTTP(S)
There is experimental support for streaming directly from HTTP or HTTPS. A URL can be passed instead of a file path:
It is possible to verify if a file is supported programmatically using helpers.is_streamable:
from pyatv.helpers import is_streamable if await is_streamable("myfile.mp3"): await atv.stream_file("myfile.mp3") else: print("File is not supported")
There are a few caveats worth knowing:
- No exception is ever raised, even when file is not found or lack of permissions
- Only valid for interface.Stream.stream_file (not interface.Stream.play_url)
- Only a basic check is made, the file might be broken and not still not playable
If you stream audio using the RAOP protocol and the device requires a password, you can set the password like this:
raop_service = atv_conf.get_service(Protocol.RAOP) raop_service.password = "test" atv = await connect(atv_conf, ...) await atv.stream.stream_file("sample.mp3")
In tvOS 10.2, Apple started to enforce a feature called “device authentication”. This requires every device that streams content via AirPlay to enter a PIN code the first time before playback is started. Once done, the user will never have to do this again. The actual feature has been available for a while but as opt-in, so it would have to be explicitly enabled. Now it is enabled by default and cannot be disabled. Devices not running tvOS (e.g. Apple TV 2nd and 3rd generation) are not affected, even though device authentication can be enabled on these devices as well.
The device authentication process is based on the Secure Remote Password protocol (SRP), with slight modifications. All the reverse engineering required for this process was made by funtax (GitHub username) and has merely been ported to python for usage in this library. Please see references at bottom of page for reference implementation.
Device Pairing in pyatv
When performing device authentication, a device identifier and a private key is required. Once authenticated, they can be used to authenticate without using a PIN code. So they must be saved and re-used whenever something is to be played.
In this library, the device identifier and private key is called AirPlay credentials and are concatenated into a string, using : as separator. An example might look like this:
D9B75D737BE2F0F1:6A26D8EB6F4AE2408757D5CA5FF9C37E96BEBB22C632426C4A02AD4FA895A85B ^ ^ Identifier Private key
The device authentication is performed via the pairing API, just like with any other protocol. New random credentials are generated by default, as long as no existing credentials are provided. So there is nothing special here.