yrriban/dcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Hardcode fewer things. Convert tabs to spaces.

Browse source
This commit is contained in:
yrriban 2025-06-12 23:47:22 -04:00
parent 2e9b8b148e
commit 97616b341d
1 changed files with 122 additions and 108 deletions
Download patch file Download diff file Expand all files Collapse all files

230
dcc/concat.py
View file

@ -9,117 +9,131 @@ import numpy as np
import wand.image import wand.image
class Concat(dcc.doom_base.Wad): class Concat(dcc.doom_base.Wad):
def get_parser(self, prog_name): def get_parser(self, prog_name):
parser = super().get_parser(prog_name) parser = super().get_parser(prog_name)
parser.add_argument("start_map") parser.add_argument("start_map")
parser.add_argument("end_map") parser.add_argument("end_map")
parser.add_argument("-n", "--nooverlay", action="store_true") parser.add_argument("-n", "--nooverlay", action="store_true")
return parser return parser
def take_action(self, parsed_args): def take_action(self, parsed_args):
logging.basicConfig() logging.basicConfig()
av.logging.set_level(av.logging.VERBOSE) av.logging.set_level(av.logging.VERBOSE)
av.logging.restore_default_callback() av.logging.restore_default_callback()
videos = self.fabricate.joinpath(parsed_args.wad).glob(f"{parsed_args.wad}_map*.mp4") videos = self.fabricate.joinpath(parsed_args.wad).glob(f"{parsed_args.wad}_map*.mp4")
output = av.open(self.fabricate.joinpath(parsed_args.wad).joinpath(f"{parsed_args.wad}_maps{parsed_args.start_map}to{parsed_args.end_map}.mp4"), "w") output = av.open(self.fabricate.joinpath(parsed_args.wad).joinpath(f"{parsed_args.wad}_maps{parsed_args.start_map}to{parsed_args.end_map}.mp4"), "w")
offset = 0 offset = 0
# We'd like to use the concat filter here and connect everything into a # We'd like to use the concat filter here and connect everything into a
# single filter graph... but it produces a "Resource temporarily # single filter graph... but it produces a "Resource temporarily
# unavailable" error when switching to inputs after the first. Presumably # unavailable" error when switching to inputs after the first. Presumably
# fixable, but it's easier to just make one graph per video and mux # fixable, but it's easier to just make one graph per video and mux
# everything together at the end. # everything together at the end.
for v in sorted(videos): for v in sorted(videos):
# TODO: Support UDoom in literally any way. # TODO: Support UDoom in literally any way.
if not (v.name >= f"{parsed_args.wad}_map{parsed_args.start_map}.mp4" and if not (v.name >= f"{parsed_args.wad}_map{parsed_args.start_map}.mp4" and
v.name <= f"{parsed_args.wad}_map{parsed_args.end_map}.mp4"): v.name <= f"{parsed_args.wad}_map{parsed_args.end_map}.mp4"):
continue continue
chunk = av.open(v) chunk = av.open(v)
ograph = av.filter.Graph() if not (len(chunk.streams.video) == 1 and len(chunk.streams.audio) == 1):
sink = ograph.add("buffersink") raise Exception(f"irregular chunk {v}: streams {chunk.streams} (expected 1 video & 1 audio)")
asink = ograph.add("abuffersink")
if not parsed_args.nooverlay: ograph = av.filter.Graph()
img = wand.image.Image(height=chunk.streams[0].height,width=chunk.streams[0].width) sink = ograph.add("buffersink")
mapstring = v.name[-6:-4] asink = ograph.add("abuffersink")
text = self._config["map_names"][f"map{mapstring}"]
dcc.text.draw_text(img, f"MAP{mapstring}: {text}", font_size=120)
img.trim(reset_coords=True)
img.border("graya(25%, 25%)", 10, 10)
img.border(dcc.config.TEXT_STROKE_COLOR, 16, 16)
# for this to work... the image needs to have a width that's a multiple
# of 8. dude whyyyyyyy
padfactor=8
img.border("transparent", padfactor, 0)
img.crop(width=img.width-img.width%padfactor, height=img.height)
text_frame = av.video.frame.VideoFrame(img.width, img.height, format="rgba")
if len(output.streams.get()) == 0: if not parsed_args.nooverlay:
# TODO: less hardcoding. img = wand.image.Image(height=chunk.streams[0].height,width=chunk.streams[0].width)
output.add_stream("h264", rate=61440) mapstring = v.name[-6:-4]
output.streams[0].extradata = copy.deepcopy(chunk.streams[0].extradata) text = self._config["map_names"][f"map{mapstring}"]
output.streams[0].height=1440 dcc.text.draw_text(img, f"MAP{mapstring}: {text}", font_size=120)
output.streams[0].width=2560 img.trim(reset_coords=True)
output.streams[0].profile="High" img.border("graya(25%, 25%)", 10, 10)
output.streams[0].qmax = chunk.streams[0].qmax img.border(dcc.config.TEXT_STROKE_COLOR, 16, 16)
output.streams[0].qmin = chunk.streams[0].qmin # for this to work... the image needs to have a width that's a multiple
output.streams[0].codec_context.gop_size=30 # of 8. dude whyyyyyyy
output.streams[0].codec_context.max_b_frames=2 padfactor=8
output.streams[0].codec_context.framerate = fractions.Fraction(60,1) img.border("transparent", padfactor, 0)
output.streams[0].codec_context.pix_fmt="yuv420p" img.crop(width=img.width-img.width%padfactor, height=img.height)
output.streams[0].codec_context.bit_rate = chunk.streams[0].codec_context.bit_rate
output.add_stream("aac", rate=48000)
output.streams[1].extradata = copy.deepcopy(output.streams[1].extradata)
output.streams[1].rate=48000
output.streams[1].bit_rate=chunk.streams[1].bit_rate
src = ograph.add_buffer(template=chunk.streams[0], time_base=chunk.streams[0].time_base)
asrc = ograph.add_abuffer(template=chunk.streams[1], time_base=chunk.streams[1].time_base)
ifade = ograph.add("fade", args="in:0:60")
iafade = ograph.add("afade", args="in:{}:48000".format(offset*48000/1000000))
ofade = ograph.add("fade", args="out:{}:60".format((chunk.duration*60/1000000)-60))
oafade = ograph.add("afade", args="out:{}:48000".format(((offset+chunk.duration)*48000/1000000)-48000))
if not parsed_args.nooverlay:
overlay = ograph.add_buffer(width=img.width, height=img.height, format="rgba", time_base=chunk.streams[0].time_base)
overlay_fo = ograph.add("fade", args="out:240:60")
overlay.link_to(overlay_fo, 0, 0)
composite = ograph.add("overlay", args="x=4:y=4")
src.link_to(composite, 0, 0)
overlay_fo.link_to(composite, 0, 1)
composite.link_to(ifade, 0, 0)
else:
src.link_to(ifade, 0, 0)
asrc.link_to(iafade, 0, 0)
ifade.link_to(ofade, 0, 0)
iafade.link_to(oafade, 0, 0)
ofade.link_to(sink, 0, 0)
oafade.link_to(asink, 0, 0)
ograph.configure() if len(output.streams.get()) == 0:
for packet in chunk.demux(): # We can't use the input stream as a template here; it doesn't
if packet.dts is None: # have everything needed to do encoding and will fail
continue # mysteriously later.
packet.dts += (offset * packet.time_base.denominator) / (packet.time_base.numerator * 1000000) vs = chunk.streams.video[0]
packet.pts += (offset * packet.time_base.denominator) / (packet.time_base.numerator * 1000000) output.add_stream("h264", rate=int(vs.time_base.denominator/vs.time_base.numerator))
if packet.stream_index == 0: # TODO: robustness output.streams[0].extradata = copy.deepcopy(vs.extradata)
for ifr in packet.decode(): output.streams[0].height=vs.height
if not parsed_args.nooverlay: output.streams[0].width=vs.width
text_frame = av.video.frame.VideoFrame(img.width, img.height, format="rgba") output.streams[0].qmax = vs.qmax
text_frame.planes[0].update(img.make_blob(format="rgba")) output.streams[0].qmin = vs.qmin
text_frame.pts = ifr.pts output.streams[0].codec_context.bit_rate = vs.codec_context.bit_rate
text_frame.dts = ifr.dts output.streams[0].codec_context.framerate = vs.base_rate
text_frame.time_base = ifr.time_base output.streams[0].codec_context.pix_fmt = vs.codec_context.pix_fmt
overlay.push(text_frame) # The following are only used for encoding and have no equivalent on the input stream.
src.push(ifr) output.streams[0].profile="High"
ofr = sink.pull() output.streams[0].codec_context.gop_size=30
for p in output.streams[packet.stream_index].encode(ofr): output.streams[0].codec_context.max_b_frames=2
output.mux(p)
else: astr = chunk.streams.audio[0]
for ifr in packet.decode(): output.add_stream("aac", rate=astr.rate)
asrc.push(ifr) output.streams[1].extradata = copy.deepcopy(astr.extradata)
ofr = asink.pull() output.streams[1].bit_rate=astr.bit_rate
for p in output.streams[packet.stream_index].encode(ofr):
output.mux(p) src = ograph.add_buffer(template=chunk.streams.video[0], time_base=chunk.streams.video[0].time_base)
offset += chunk.duration asrc = ograph.add_abuffer(template=chunk.streams.audio[0], time_base=chunk.streams.audio[0].time_base)
chunk.close() # TODO: video fades are absolute relative to the input video; audio
output.close() # fades need to have their timestamps offset by the position in the
# final video. Clarify if this is really necessary.
frame_rate = chunk.streams.video[0].base_rate
sample_rate = chunk.streams.audio[0].rate
ifade = ograph.add("fade", args="in:0:{}".format(frame_rate))
ofade = ograph.add("fade", args="out:{}:{}".format((chunk.duration*frame_rate/1000000)-frame_rate, frame_rate))
iafade = ograph.add("afade", args="in:{}:{}".format(offset*sample_rate/1000000, sample_rate))
oafade = ograph.add("afade", args="out:{}:{}".format(((offset+chunk.duration)*sample_rate/1000000)-sample_rate, sample_rate))
if not parsed_args.nooverlay:
overlay = ograph.add_buffer(width=img.width, height=img.height, format="rgba", time_base=chunk.streams[0].time_base)
overlay_fo = ograph.add("fade", args="out:{}:{}".format(4*frame_rate, frame_rate))
overlay.link_to(overlay_fo, 0, 0)
composite = ograph.add("overlay", args="x=4:y=4")
src.link_to(composite, 0, 0)
overlay_fo.link_to(composite, 0, 1)
composite.link_to(ifade, 0, 0)
else:
src.link_to(ifade, 0, 0)
asrc.link_to(iafade, 0, 0)
ifade.link_to(ofade, 0, 0)
iafade.link_to(oafade, 0, 0)
ofade.link_to(sink, 0, 0)
oafade.link_to(asink, 0, 0)
ograph.configure()
for packet in chunk.demux():
if packet.dts is None:
continue
packet.dts += (offset * packet.time_base.denominator) / (packet.time_base.numerator * 1000000)
packet.pts += (offset * packet.time_base.denominator) / (packet.time_base.numerator * 1000000)
if packet.stream == chunk.streams.video[0]:
for ifr in packet.decode():
if not parsed_args.nooverlay:
text_frame = av.video.frame.VideoFrame(img.width, img.height, format="rgba")
text_frame.planes[0].update(img.make_blob(format="rgba"))
text_frame.pts = ifr.pts
text_frame.dts = ifr.dts
text_frame.time_base = ifr.time_base
overlay.push(text_frame)
src.push(ifr)
ofr = sink.pull()
for p in output.streams[packet.stream_index].encode(ofr):
output.mux(p)
else:
for ifr in packet.decode():
asrc.push(ifr)
ofr = asink.pull()
for p in output.streams[packet.stream_index].encode(ofr):
output.mux(p)
offset += chunk.duration
chunk.close()
output.close()