yt_common: automation: vfr/timecodes support

Princess Principal/pripri_common/filter.py

@@ -5,11 +5,11 @@ from typing import Any, Dict, Optional
 
 from yt_common.antialiasing import sraa_clamp, mask_strong
 from yt_common.data import FSRCNNX
-from yt_common.denoise import bm3d
 
 from G41Fun import MaskedDHA
 from awsmfunc import bbmod
 from kagefunc import retinex_edgemask
+from lvsfunc.denoise import bm3d
 from lvsfunc.kernels import Bicubic, Kernel
 from lvsfunc.misc import replace_ranges, scale_thresh
 from lvsfunc.scale import descale as ldescale

Vivy/vivy_common/config.py

@@ -14,10 +14,10 @@ TITLE_LONG: str = f"{TITLE} - Fluorite Eye's Song"
 RESOLUTION: int = 1080
 DATAPATH: str = os.path.dirname(__file__)
 
-WAKA_RU_FILENAME: str = f"{TITLE}_{{epnum:02d}}_RU_HD.mp4"
-WAKA_FR_FILENAME: str = f"{TITLE}_{{epnum:02d}}_FR_HD.mp4"
-WAKA_DE_FILENAME: str = f"{TITLE} - Fluorite Eyes Song E{{epnum:02d}} [1080p][AAC][JapDub][GerSub][Web-DL].mkv"
-AMAZON_FILENAME: str = "{title_long} - {epnum:02d} (Amazon Prime VBR {resolution}p).mkv"
+WAKA_RU_FILENAME: str = f"{TITLE}_{{epnum:s}}_RU_HD.mp4"
+WAKA_FR_FILENAME: str = f"{TITLE}_{{epnum:s}}_FR_HD.mp4"
+WAKA_DE_FILENAME: str = f"{TITLE} - Fluorite Eyes Song E{{epnum:s}} [1080p][AAC][JapDub][GerSub][Web-DL].mkv"
+AMAZON_FILENAME: str = "{title_long} - {epnum:s} (Amazon Prime VBR {resolution}p).mkv"
 
 
 core = vs.core
@@ -37,13 +37,13 @@ class VivyConfig(Config):
 class VivySource(source.FunimationSource):
     def get_amazon(self) -> vs.VideoNode:
         # ep1-3 have good funi video, let's just use that
-        if self.config.epnum < 4:
+        if int(self.config.desc) < 4:
             raise FileNotFoundError()
         if not os.path.isfile(self.config.format_filename(AMAZON_FILENAME_VBR)):
             log.warn("Amazon not found, falling back to Funimation")
             raise FileNotFoundError()
         log.success("Found Amazon video")
-        return core.ffms2.Source(self.config.format_filename(AMAZON_FILENAME_VBR))
+        return self._open(self.config.format_filename(AMAZON_FILENAME_VBR))
 
     def get_waka_filenames(self) -> List[str]:
         return [self.config.format_filename(f) for f in [

Vivy/vivy_common/filter.py

@@ -6,6 +6,7 @@ import vardefunc as vdf
 from awsmfunc import bbmod
 from debandshit import f3kbilateral
 from lvsfunc.aa import upscaled_sraa
+from lvsfunc.denoise import bm3d
 from lvsfunc.kernels import Bicubic
 from lvsfunc.misc import replace_ranges
 from lvsfunc.types import Range
@@ -13,7 +14,6 @@ from typing import List, Optional, Union
 
 from yt_common import antialiasing
 from yt_common.data import FSRCNNX
-from yt_common.denoise import bm3d
 from yt_common.deband import morpho_mask
 
 import vsutil

yt_common/yt_common/__init__.py

@@ -1 +1 @@
-from . import antialiasing, automation, config, data, deband, denoise, logging, source  # noqa: F401
+from . import antialiasing, automation, config, data, deband, logging, source  # noqa: F401

yt_common/yt_common/automation.py

@@ -8,7 +8,9 @@ import shutil
 import string
 import subprocess
 
-from typing import Any, BinaryIO, Callable, List, Optional, Sequence, cast
+from lvsfunc.render import clip_async_render
+
+from typing import Any, BinaryIO, Callable, List, Optional, Sequence, Tuple, cast
 
 from .config import Config
 from .logging import log
@@ -49,14 +51,16 @@ class Encoder():
 
         self._get_encoder_settings(settings_path)
 
-    def encode(self, clip: vs.VideoNode, filename: str, start: int = 0, end: int = 0) -> str:
+    def encode(self, clip: vs.VideoNode, filename: str, start: int = 0, end: int = 0,
+               timecode_file: Optional[str] = None, want_timecodes: bool = False) -> Tuple[str, List[float]]:
         end = end if end != 0 else clip.num_frames
+        want_timecodes = True if timecode_file else want_timecodes
 
         outfile = self.out_template.format(filename=filename)
 
         if os.path.isfile(outfile) and not self.force:
             log.warn("Existing output detected, skipping encode!")
-            return outfile
+            return outfile, []
 
         params = [p.format(frames=end-start, filename=filename) for p in self.params]
 
@@ -71,7 +75,13 @@ class Encoder():
         # signal.signal(signal.SIGINT, forward_to_proc)
         # turns out this didn't work out the way i had hoped
 
-        clip[start:end].output(cast(BinaryIO, process.stdin), y4m=True)
+        # use the python renderer only if we need timecodes because it's slower
+        timecodes: List[float] = []
+        if want_timecodes:
+            timecode_io = open(timecode_file, "w") if timecode_file else None
+            timecodes = clip_async_render(clip[start:end], cast(BinaryIO, process.stdin), timecode_io)
+        else:
+            clip[start:end].output(cast(BinaryIO, process.stdin), y4m=True)
         process.communicate()
 
         # vapoursynth should handle this itself but just in case
@@ -81,7 +91,7 @@ class Encoder():
 
         log.success("--- ENCODE FINISHED ---")
         self.cleanup.append(outfile)
-        return outfile
+        return outfile, timecodes
 
     def _get_encoder_settings(self, settings_path: str) -> None:
         with open(settings_path, "r") as settings:
@@ -116,7 +126,7 @@ class AudioGetter():
         trims = self.src.get_audio()
         if not trims or len(trims) > 1:
             raise NotImplementedError("Please implement multifile trimming")
-        audio_cut = acsuite.eztrim(trims[0].path, cast(acsuite.types.Trim, trims[0].trim), streams=0)[0]
+        audio_cut = acsuite.eztrim(trims[0].path, trims[0].trim or (0, None), streams=0)[0]
         self.cleanup.append(audio_cut)
 
         if ftrim:
@@ -155,6 +165,8 @@ class SelfRunner():
 
     video_file: str
     audio_file: str
+    timecodes: List[float]
+    timecode_file: Optional[str]
 
     encoder: Encoder
     audio: AudioGetter
@@ -243,8 +255,16 @@ class SelfRunner():
             raise FileNotFoundError(f"Failed to find {settings_path}!")
 
         self.encoder = Encoder(settings_path, args.encoder, args.force)
-        self.video_file = self.encoder.encode(self.clip, f"{self.config.desc}_{self.suffix}_{start}_{end}",
-                                              start, end)
+        # we only want to generate timecodes if vfr, otherwise we can just calculate them
+        self.timecode_file = f"{self.config.desc}_{self.suffix}_{start}_{end}_timecodes.txt" \
+            if self.clip.fps_den == 0 else None
+        self.video_file, self.timecodes = self.encoder.encode(self.clip,
+                                                              f"{self.config.desc}_{self.suffix}_{start}_{end}",
+                                                              start, end, self.timecode_file)
+
+        # calculate timecodes if cfr and we didn't generate any (we shouldn'tve)
+        self.timecodes = [round(float(1e9*f*(1/self.clip.fps)))/1e9 for f in range(0, self.clip.num_frames + 1)] \
+            if self.clip.fps_den != 0 and len(self.timecodes) == 0 else self.timecodes
 
         self._do_audio(start, end, audio_codec)
 
@@ -280,11 +300,13 @@ class SelfRunner():
             self.audio_file = out_name
 
     def _do_mux(self, name: str, chapters: bool = True) -> int:
+        tcargs = ["--timecodes", f"0:{self.timecode_file}"] if self.timecode_file else []
         mkvtoolnix_args = [
             "mkvmerge",
             "--output", name,
             "--no-chapters", "--no-track-tags", "--no-global-tags", "--track-name", "0:",
             "--default-track", "0:yes",
+        ] + tcargs + [
             "(", self.video_file, ")",
             "--no-chapters", "--no-track-tags", "--no-global-tags", "--track-name", "0:",
             "--default-track", "0:yes", "--language", "0:jpn",

yt_common/yt_common/denoise.py

@@ -1,65 +0,0 @@
-import vapoursynth as vs
-from vsutil import iterate, Range, get_y
-from typing import Any, Dict, List, Optional, Union
-
-
-core = vs.core
-
-
-# always uses float OPP
-# mvsfunc is an unreadable mess and i was having bm3d issues so i rewrote the wrapper
-# turns out, it's still an unreadable mess
-def bm3d(clip: vs.VideoNode, matrix_s: str = "709", sigma: Union[float, List[float]] = 0.75,
-         radius: Union[int, List[int], None] = None, ref: Optional[vs.VideoNode] = None,
-         pre: Optional[vs.VideoNode] = None, refine: int = 1,
-         basic_args: Dict[str, Any] = {}, final_args: Dict[str, Any] = {}) -> vs.VideoNode:
-    assert clip.format is not None
-    isGray = clip.format.color_family == vs.GRAY
-
-    def to_opp(clip: vs.VideoNode) -> vs.VideoNode:
-        return clip.resize.Bicubic(format=vs.RGBS, matrix_in_s=matrix_s).bm3d.RGB2OPP(sample=1)
-
-    def to_fullgray(clip: vs.VideoNode) -> vs.VideoNode:
-        return get_y(clip).resize.Point(format=vs.GRAYS, range_in=Range.LIMITED, range=Range.FULL)
-
-    sigmal = [sigma] * 3 if not isinstance(sigma, list) else sigma + [sigma[-1]]*(3-len(sigma))
-    sigmal = [sigmal[0], 0, 0] if isGray else sigmal
-    isGray = True if sigmal[1] == 0 and sigmal[2] == 0 else isGray
-    if len(sigmal) != 3:
-        raise ValueError("bm3d: 'invalid number of sigma parameters supplied'")
-    radiusl = [0, 0] if radius is None else [radius] * 2 if not isinstance(radius, list) \
-        else radius + [radius[-1]]*(2-len(radius))
-    if len(radiusl) != 2:
-        raise ValueError("bm3d: 'invalid number or radius parameters supplied'")
-
-    if sigmal[0] == 0 and sigmal[1] == 0 and sigmal[2] == 0:
-        return clip
-
-    pre = pre if pre is None else to_opp(pre) if not isGray else to_fullgray(pre)
-
-    def basic(clip: vs.VideoNode) -> vs.VideoNode:
-        return clip.bm3d.Basic(sigma=sigmal, ref=pre, matrix=100, **basic_args) if radiusl[0] < 1 \
-            else clip.bm3d.VBasic(sigma=sigmal, ref=pre, radius=radiusl[0], matrix=100, **basic_args) \
-            .bm3d.VAggregate(radius=radiusl[0], sample=1)
-
-    clip_in = to_opp(clip) if not isGray else to_fullgray(clip)
-    refv = basic(clip_in) if ref is None else to_opp(ref) if not isGray else to_fullgray(ref)
-
-    def final(clip: vs.VideoNode) -> vs.VideoNode:
-        return clip.bm3d.Final(sigma=sigmal, ref=refv, matrix=100, **final_args) if radiusl[1] < 1 \
-            else clip.bm3d.VFinal(sigma=sigmal, ref=refv, radius=radiusl[1], matrix=100, **final_args) \
-            .bm3d.VAggregate(radius=radiusl[1], sample=1)
-
-    den = iterate(clip_in, final, refine)
-
-    # boil everything back down to whatever input we had
-    den = den.bm3d.OPP2RGB(sample=1).resize.Bicubic(format=clip.format.id, matrix_s=matrix_s) if not isGray \
-        else den.resize.Point(format=clip.format.replace(color_family=vs.GRAY, subsampling_w=0, subsampling_h=0).id,
-                              range_in=Range.FULL, range=Range.LIMITED)
-    # merge source chroma if it exists and we didn't denoise it
-    den = core.std.ShufflePlanes([den, clip], planes=[0, 1, 2], colorfamily=vs.YUV) \
-        if isGray and clip.format.color_family == vs.YUV else den
-    # sub clip luma back in if we only denoised chroma
-    den = den if sigmal[0] != 0 else core.std.ShufflePlanes([clip, den], planes=[0, 1, 2], colorfamily=vs.YUV)
-
-    return den