Closed fraroas closed 1 year ago
It's not turned on, because a lot of stuff doesn't respect auto-return in SCTE-35 Cues, I need to look at the code and then I tell you where it should be called.
x9k3 will also read cues from a sidecar file, you don't have to inject them. threefive can generate cues
mk_splice_insert(event_id, pts, duration=None)
mk_cue returns a Cue
with a Splice Insert.
splice_event_id = event_id
If duration is NOT set,
out_of_network_indicator False
time_specified_flag False
duration_flag False
splice_immediate_flag True
if duration IS set:
out_of_network_indicator True
time_specified_flag True
duration_flag True
splice_immediate_flag False
break_auto_return True
break_duration duration
pts_time pts
a@debian:~/cuei$ pypy3
>>>> from threefive.encode import mk_splice_insert
>>>> evnt_id= 9
>>>> pts = 29053.123456
>>>> duration = 300
>>>> cue =mk_splice_insert(evnt_id,pts,duration)
>>>> cue.encode()
'/DAlAAAAAAAAAP/wFAUAAAAJf+/+m9pkt/4Bm/zAAAkAAAAA5ftzmA=='
>>>> cue.encode_as_hex()
'0xfc302500000000000000fff01405000000097feffe9bda64b7fe019bfcc0000900000000e5fb7398'
>>>> cue.encode_as_int()
2104181392760166021170929920497819646350710275162000838308971042240028739408705486837725464851352
python3 -m pip install --upgrade x9k3 threefive new_reader
and auto return should work.
Thanks a lot for this quick fix! Great!
However, there is a new issue with "immediate" splice requests, when no pts is given: File "x9k3/x9k3.py", line 534, in _auto_return pts = cmd.pts_time + cmd.break_duration TypeError: unsupported operand type(s) for +: 'NoneType' and 'float'
I guess the pts_default should be set to current pts (or of the next packet) instead to None if there is no pts_time given in the section.
tsduck splice xml looks like above, which is
in line 534 change cmd.pts_time to self.scte35.cue_time
pts = self.scte35.cue_time + cmd.break_duration
see if that works
no, the self.scte35.cue_time is still empty (None). I'll do more debugging tomorrow. In encode.py, line 98 I had to cast the pts to string. This correctly prints out the pts when the slice event was received.
/tmp/seg0.ts start: 32514.760000 duration: 3.600000 stream diff: 0.069976 /tmp/seg1.ts start: 32518.360000 duration: 3.840000 stream diff: 0.067004
(*spliceinject is from tsduck, piping to | python3 x9k3.py -d -o /tmp -t 2 -T x_cue The complete command line is tsp -v --add-input-stuffing 1/20 -I ip 224.0.0.1:6600 -P pmt --service 0x1 --add-programinfo-id 0x43554549 --add-pid 600/0x86 -P spliceinject --udp 4444 --service 0x1 -P filter --negate --pid 0x1FFF | python3 x9k3.py -d -o /tmp -t 2 -T x_cue And receiving a regular TS via multicast. Inject the XML by cat splice.xml > /dev/udp/127.0.0.1/4444
That command is complex and not needed.
x9k3 has multicast support.
x9k3 can read cues from a sidecar file.
grab the latest threefive and x9k3
pypy3 -mpip install --upgrade threefive x9k3
Generate a Spice Insert with splice_immediate_flag and break auto return
from threefive.encode import mk_splice_insert
evnt_id= 9
pts =None # None for Splice Immediate
duration = 300
insert_pts = 1234.567890
cue =mk_splice_insert(evnt_id,pts,duration)
with open("sidecar.txt, "a") as scf:
scf.write(f"{insert_pts},{cue.encode()} ")
x9k3 -i udp://@224.0.0.1:6600 -s sidecar.txt -T x_cue -o /tmp
Did it work for you?
~/build/clean/x9k3$ pypy3
Python 3.8.13 (7.3.9+dfsg-5, Oct 30 2022, 09:55:31)
[PyPy 7.3.9 with GCC 12.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
from threefive.encode import mk_splice_insert
evt_id, pts, dur = 9, None, 13.4
cue = mk_splice_insert(evt_id,pts,dur).encode()
with open("sidecar.txt","w") as sidecar: .... sidecar.write(f'113.456789,{cue}') ....
59I ran this x9k3 command ```lua a@debian:~/build/clean/x9k3$ x9k3 -i la-slim.ts -s sidecar.txt
Loading from Sidecar File 113.456789 /DAhAAAAAAAAAP/wEAUAAAAJf78A/gASZvAACQAAAACokv3z ./seg0.ts start: 102.506478 duration: 2.126855 ./seg1.ts start: 104.633333 duration: 2.000000 ./seg2.ts start: 106.633333 duration: 2.000000 ./seg3.ts start: 108.633333 duration: 2.000000 ./seg4.ts start: 110.633333 duration: 2.000000
Loading from Sidecar File 113.456789 /DAhAAAAAAAAAP/wEAUAAAAJf78A/gASZvAACQAAAACokv3z
Loading from Sidecar File 113.456789 /DAhAAAAAAAAAP/wEAUAAAAJf78A/gASZvAACQAAAACokv3z
./seg5.ts start: 112.633333 duration: 2.000000
./seg6.ts start: 114.633333 duration: 2.000000
./seg7.ts start: 116.633333 duration: 2.000000
./seg8.ts start: 118.633333 duration: 2.000000
./seg9.ts start: 120.633333 duration: 2.000000
./seg10.ts start: 122.633333 duration: 2.000000
./seg11.ts start: 124.633333 duration: 2.000000
./seg12.ts start: 126.633333 duration: 2.000000 ./seg13.ts start: 128.633333 duration: 2.000000 ./seg14.ts start: 130.633333 duration: 2.000000 ./seg15.ts start: 132.633333 duration: 2.033334 ./seg16.ts start: 134.666667 duration: 2.000000 ./seg17.ts start: 136.666667 duration: 2.000000 ./seg18.ts start: 138.666667 duration: 2.000000 ./seg19.ts start: 140.666667 duration: 2.000000 ./seg20.ts start: 142.666667 duration: 2.000000 ./seg21.ts start: 144.666667 duration: 2.000000 ./seg22.ts start: 146.666667 duration: 2.000000 ./seg23.ts start: 148.666667 duration: 2.000000 ./seg24.ts start: 150.666667 duration: 2.000000
I can confirm that auto_return works for "immediate" cues, injected by sidecar created by threefive.encode. Having no sidecar but using a recorded ts with markers leads to File "/home/frank/src/x9k3/x9k3.py", line 540, in _auto_return if [pts, cue] not in self.sidecar:
I'll open another issue regarding the alignment of HLS ad markers to GOP boundaries.
changing x9k3.py line 540 to if self.sidecar and [pts, cue] not in self.sidecar: fixes the error above, but then the auto_return still doesn't work:
/tmp/seg155.ts start: 19309.086222 duration: 4.040000 pts for cue in: 19407.326222 /tmp/seg156.ts start: 19313.126222 duration: 4.000000
/tmp/seg157.ts start: 19317.126222 duration: 4.000000
/tmp/seg158.ts start: 19321.126222 duration: 4.080000
/tmp/seg159.ts start: 19325.206222 duration: 4.080000 ...
/tmp/seg178.ts start: 19401.646222 duration: 4.000000
/tmp/seg179.ts start: 19405.646222 duration: 4.040000
/tmp/seg180.ts start: 19409.686222 duration: 4.000000
ah, I see. You're adding the automatically created cue-in to the sidecar data. But if there is no sidecar, the auto-return won't work. We should store the cue-in locally as the next cue msg.
You are confusing the sidecar file with the sidecar object, two different things. The sidecar file is read and loaded into the sidecar object.
I am going to pull out the auto return stuff, it is causing too many problems. I am just cant support auto return right now. I am way too busy to spend any more time on this right now. If you get how you want it , send me a patch and I will apply it.
I am in the middle of re-writing x9k3, and I need to finish that first, and then add auto-return
Its not line 540, its line 539 that needs to be changed to
if [pts, b64] not in self.sidecar: # line 539
self.sidecar.append([pts, b64]) # line 540
pip up
pypy3 -mpip install --upgrade x9k3
work perfectly for me:
a@debian:~/build/clean/x9k3$ x9k3 -i ../la-slim.ts -s sidecar.txt
./seg0.ts start: 102.506478 duration: 2.126855
./seg1.ts start: 104.633333 duration: 2.000000
./seg2.ts start: 106.633333 duration: 2.000000
./seg3.ts start: 108.633333 duration: 2.000000
./seg4.ts start: 110.633333 duration: 2.000000
./seg5.ts start: 112.633333 duration: 2.000000
#EXT-X-CUE-OUT:13.4
./seg6.ts start: 114.633333 duration: 2.000000
#EXT-X-CUE-OUT-CONT:2.000/13.4
./seg7.ts start: 116.633333 duration: 2.000000
#EXT-X-CUE-OUT-CONT:4.000/13.4
./seg8.ts start: 118.633333 duration: 2.000000
#EXT-X-CUE-OUT-CONT:6.000/13.4
./seg9.ts start: 120.633333 duration: 2.000000
#EXT-X-CUE-OUT-CONT:8.000/13.4
./seg10.ts start: 122.633333 duration: 2.000000
#EXT-X-CUE-OUT-CONT:10.000/13.4
./seg11.ts start: 124.633333 duration: 2.000000
#EXT-X-CUE-OUT-CONT:12.000/13.4
./seg12.ts start: 126.633333 duration: 2.000000
#EXT-X-CUE-IN
./seg13.ts start: 128.633333 duration: 2.000000
./seg14.ts start: 130.633333 duration: 2.000000
./seg15.ts start: 132.633333 duration: 2.033334
./seg16.ts start: 134.666667 duration: 2.000000
./seg17.ts start: 136.666667 duration: 2.000000
./seg18.ts start: 138.666667 duration: 2.000000
./seg19.ts start: 140.666667 duration: 2.000000
./seg20.ts start: 142.666667 duration: 2.000000
./seg21.ts start: 144.666667 duration: 2.000000
./seg22.ts start: 146.666667 duration: 2.000000
./seg23.ts start: 148.666667 duration: 2.000000
./seg24.ts start: 150.666667 duration: 2.000000
a@debian:~/build/clean/x9k3$
If it doesnt work for you, it will have to wait until I finish my big update.
make sure your cues are on iframes.
I have a tool for that as well called iframes
a@debian:~/build/clean/x9k3$ iframes ../la-slim.ts
102.966667
103.3
103.633333
103.966667
104.3
104.633333
104.966667
105.3
105.633333
105.966667
106.3
106.633333
106.966667
107.3
107.633333
107.966667
108.3
108.633333
108.966667
109.3
109.633333
109.966667
110.3
110.633333
110.966667
111.3
111.633333
111.966667
112.3
112.633333
112.966667
113.3
113.633333
113.966667
114.3
114.633333
114.966667
115.3
115.633333
115.966667
116.3
116.633333
116.966667
117.3
117.633333
117.966667
118.3
118.633333
118.966667
119.3
119.633333
119.966667
120.3
120.633333
120.966667
121.3
121.633333
121.966667
122.3
122.633333
122.966667
123.3
123.633333
123.966667
124.3
124.633333
124.966667
125.3
125.633333
125.966667
126.3
126.633333
126.966667
127.3
127.633333
127.966667
128.3
128.633333
128.966667
129.3
129.633333
129.966667
130.3
130.633333
130.966667
131.3
131.633333
131.966667
132.3
132.633333
133.0
133.333333
133.666667
134.0
134.333333
134.666667
135.0
135.333333
135.666667
136.0
136.333333
136.666667
137.0
137.333333
137.666667
138.0
138.333333
138.666667
139.0
139.333333
139.666667
140.0
140.333333
140.666667
141.0
141.333333
141.666667
142.0
142.333333
142.666667
143.0
143.333333
143.666667
144.0
144.333333
144.666667
145.0
145.333333
145.666667
146.0
146.333333
146.666667
147.0
147.333333
147.666667
148.0
148.333333
148.666667
149.0
149.333333
149.666667
150.0
150.333333
150.666667
151.0
151.333333
151.666667
152.0
152.333333
152.666667
iframes matches up with ffprobe,
102.966667 ,
103.300000 ,
103.633333 ,
103.966667 ,
104.300000 ,
104.633333 ,
104.966667 ,
105.300000 ,
105.633333 ,
105.966667 ,
106.300000 ,
106.633333 ,
106.966667 ,
107.300000 ,
107.633333 ,
107.966667 ,
108.300000 ,
108.633333 ,
108.966667 ,
109.300000 ,
109.633333 ,
109.966667 ,
110.300000 ,
110.633333 ,
110.966667 ,
111.300000 ,
111.633333 ,
111.966667 ,
112.300000 ,
112.633333 ,
112.966667 ,
113.300000 ,
113.633333 ,
113.966667 ,
114.300000 ,
114.633333 ,
114.966667 ,
115.300000 ,
115.633333 ,
115.966667 ,
116.300000 ,
116.633333 ,
116.966667 ,
117.300000 ,
117.633333 ,
117.966667 ,
118.300000 ,
118.633333 ,
118.966667 ,
119.300000 ,
119.633333 ,
119.966667 ,
120.300000 ,
120.633333 ,
120.966667 ,
121.300000 ,
121.633333 ,
121.966667 ,
122.300000 ,
122.633333 ,
122.966667 ,
123.300000 ,
123.633333 ,
123.966667 ,
124.300000 ,
124.633333 ,
124.966667 ,
125.300000 ,
125.633333 ,
125.966667 ,
126.300000 ,
126.633333 ,
126.966667 ,
127.300000 ,
127.633333 ,
127.966667 ,
128.300000 ,
128.633333 ,
128.966667 ,
129.300000 ,
129.633333 ,
129.966667 ,
130.300000 ,
130.633333 ,
130.966667 ,
131.300000 ,
131.633333 ,
131.966667 ,
132.300000 ,
132.633333 ,
133.000000 ,
133.333333 ,
133.666667 ,
134.000000 ,
134.333333 ,
134.666667 ,
135.000000 ,
135.333333 ,
135.666667 ,
136.000000 ,
136.333333 ,
136.666667 ,
137.000000 ,
137.333333 ,
137.666667 ,
138.000000 ,
138.333333 ,
138.666667 ,
139.000000 ,
139.333333 ,
139.666667 ,
140.000000 ,
140.333333 ,
140.666667 ,
141.000000 ,
141.333333 ,
141.666667 ,
142.000000 ,
142.333333 ,
142.666667 ,
143.000000 ,
143.333333 ,
143.666667 ,
144.000000 ,
144.333333 ,
144.666667 ,
145.000000 ,
145.333333 ,
145.666667 ,
146.000000 ,
146.333333 ,
146.666667 ,
147.000000 ,
147.333333 ,
147.666667 ,
148.000000 ,
148.333333 ,
148.666667 ,
149.000000 ,
149.333333 ,
149.666667 ,
150.000000 ,
150.333333 ,
150.666667 ,
151.000000 ,
151.333333 ,
151.666667 ,
152.000000 ,
152.333333 ,
152.666667 ,
the is nothing wrong with the iframe detection in x9k3.
I have auto-return working very well in the rewrite, probably be finished with it this week. If you having problems with iframe detection, try using shulga mode, especially if youre using mpeg2 codecs, it uses random access indicator flags.
Check out v.0.1.77 auto-return is working well.
Thanks! Will have a look tomorrow.
I'm using tsduck to inject SCTE35 Markers with ...
... However, the ad breaks runs forever:
EXT-X-CUE-OUT-CONT:825.560/23.888889
EXTINF:3.840,
seg1796.ts
EXT-X-CUE-OUT-CONT:829.400/23.888889
EXTINF:3.840,
seg1797.ts
EXT-X-CUE-OUT-CONT:833.240/23.888889
EXTINF:3.840,
I'm a total python newbie, but looking at the code, the _auto_return is never called / tested. Where should I insert the auto-return test?
Great tool, btw! Adds the last missing piece to the ffmpeg / tsduck eco system.