camera support

[wiomoc]

No description provided.

[jlusiardi]

hey, nice work!

I can perform a pairing with my iPhone using this demoserver.json

{
  "accessory_ltpk": "8bbaf1ee0e9b6669535c19355cc2e9eeec448a04c105c891dd851e88cb8674c8",
  "accessory_ltsk": "5a15ed1673db599a3224f565829f341f5a8564051149c1e1a9cd59f6fade80a3",
  "accessory_pairing_id": "12:00:00:00:01:0F",
  "accessory_pin": "031-45-154",
  "c#": 12,
  "category": "IP Camera",
  "host_ip": "192.168.178.XXX",
  "host_port": 51841,
  "name": "Camera",
  "peers": {
    "9C0C94F4-6A93-4067-993D-CBEE2D6E4006": {
      "admin": true,
      "key": "fb491bb43ca3240fac077b3c7f1055a7912451983a092bb0d2c26d263d8a51fa"
    }
  },
  "unsuccessful_tries": 0
}

but it never tries to start the ffmpeg part (which will fail anyway because i will have to adjust at leas the cam device).

[wiomoc]

Good question, because the specs are also stating in table 14-2:

Little endian integer. This must use the minimum number of bytes to encode the integer.

[jlusiardi]

Good point, can we check what an iOS device is sending/expecting? As stated at least one 'real' accessory uses 2 bytes here.

[wiomoc]

The Apple Homekit Accessory Simulator has following value for supported-audio-configuration,
the codec types are one byte long

00000000: 0111 0101 *00*02 0c01 0101 0201 0002 0101  ................
00000010: 0301 0001 1101 *01*01 020c 0101 0102 0100  ................  
00000020: 0201 0103 0100 0111 0101 *02*02 0c01 0101  ................
00000030: 0201 0003 0101 0301 0201 1701 01*03* 0212  ................
00000040: 0101 0102 0100 0201 0103 0100 0301 0103  ................
00000050: 0102 0201 00                             .....`

Strange is the fact that codec types 0 and 1 do not appear in the specification

[Jc2k]

Theres some relevant discussion that might be useful here.

There are some additional codecs that were found in the homekit binaries but not in the spec, as you noticed with the simulator:

  case 0:
    pcVar2 = &cf_HMDAudioCodecGroupTypePCMU;
    break;
  case 1:
    pcVar2 = &cf_HMDAudioCodecGroupTypePCMA;
    break;
  case 2:
    pcVar2 = &cf_HMDAudioCodecGroupTypeAACELD;
    break;
  case 3:
    pcVar2 = &cf_HMDAudioCodecGroupTypeOpus;
    break;
  case 4:
    pcVar2 = &cf_HMDAudioCodecGroupTypeMSBC;
    break;
  case 5:
    pcVar2 = &cf_HMDAudioCodecGroupTypeAMR;
    break;
  case 6:
    pcVar2 = &cf_HMDAudioCodecGroupTypeAMRWB;
    break;

Theres also some discussion around how e.g. opus is problematic.

Someone has also figured out how to make secure video work here.

[jlusiardi]

@wiomoc I started merging your branch under wiomoc_camera and an own branch camera. There I will add some commands to work with your democamserver and the circle2 camera. This ffmpeg stuff is pretty complicated for figuring out the parameters 😅

[jlusiardi]

@wiomoc i am not fully sure if i understand TLVItem properly and how it encodes lists of elements.

Take 9.105 Supported Video Stream Configuration for example. We can have multiple instances of Video Attributes to offer multiple resolutions and frame rates.

The code from the democameraserver returns the following data for 2 entries:

[[3, bytearray(b'\x01\x02\x80\x07\x02\x028\x04\x03\x01\x1e\x01\x02@\x01\x02\x01\xf0\x03\x01\x0f')]]

But the Circle2 camera returns:

[[3, bytearray(b'\x01\x02\x00\x05\x02\x02\xd0\x02\x03\x01\x1e')], [255, bytearray(b'')], [3, bytearray(b'\x01\x02\x80\x02\x02\x02h\x01\x03\x01\x1e')]]

I think this is consistent with "An IP camera accessory supporting encoding video at different resolutions must include multiple instances of Video attributes TLV." (page 221 / Spec Revision 2) and "There may be multiple, separate TLV items of the same type if separated by a TLV item of a different type." (page 254 / Spec Revision 2). Page 51 / Table 5-6 lists kTLVType_Separator = 0xff with a length of 0 as separator. Is this possible to achieve with your code?

[jlusiardi]

Hey @wiomoc,

i tried to pull out the whole TLV generation to a module of itself: https://github.com/jlusiardi/tlv8_python

This might simplify using TLV8 and I think I can code some examples how to use it in the camera area of the code.

Can you please have a look on the module?