Skip to main content

Ffmpeg wrapper for transcoding between video formats with an emphasis on maintaining quality and color depth in video production pipelines

Project description

vtcff

🚧 This project is a draft and is not intended to be used by anyone.

vtcff is a library for transcoding between video formats with an emphasis on maintaining quality and color depth in video production pipelines. In studio tasks, video seconds take up gigabytes and quality compromises are least desirable.

vtcff tends to maximize quality, sacrificing speed and disk space.

vtcff is actually a wrapper for ffmpeg – the least intuitive video tool ever created.

Install

$ pip3 install vtcff

This command will install the package, but not ffmpeg itself.

other options

Install pre-release from GitHub:

$ pip3 install git+https://github.com/rtmigo/vtcff_py@staging#egg=vtcff

Basic example

import subprocess
from vtcff import FfmpegCommand, Scale, Transpose, Hevc

cmd = FfmpegCommand()

cmd.src_file = '/path/to/source.mov'
cmd.dst_file = '/path/to/target.mov'

# set set some filters
cmd.scale = Scale(1920, 1080)
cmd.transpose = Transpose.CLOCKWISE

# set compression format
cmd.dst_codec_video = Hevc(mbps=100)

# run command
subprocess.check_call(list(cmd))

Getting the generated arguments

import subprocess, os
from vtcff import FfmpegCommand, Prores

cmd = FfmpegCommand()
cmd.src_file = 'source.mov'
cmd.dst_file = 'target.mov'
cmd.dst_codec_video = Prores()

print(str(cmd))
# ffmpeg -i source.mov -codec:v prores_ks target.mov

print(list(cmd))
# ['ffmpeg', '-i', 'source.mov', '-codec:v', 'prores_ks', 'target.mov']

# running in different ways:
os.system(str(cmd))
subprocess.run(list(cmd))

Custom arguments

The object allows you to manually specify ffmpeg arguments. Arguments given in this way take precedence over arguments generated by the object.

from vtcff import FfmpegCommand

cmd = FfmpegCommand()

# set arguments as string
cmd.custom.video.string = "-codec:v libx265"
cmd.custom.video.string += "-x265-params lossless=1"

# or as list
cmd.custom.video.list = ["-codec:v", "libx265"]
cmd.custom.video.list.extend(["-x265-params", "lossless=1"])

The cmd.custom contains four fields, that can be modified independently.

Arguments to be inserted before -i source:

  • custom.before_i

Arguments to be inserted after -i source:

  • custom.after_i
  • custom.video
  • custom.audio

zscale vs scale

ffmpeg has two video filters for color and frame size conversions:

By default, vtcff uses zscale. Sometimes it may lead to error "no path between colorspaces". If other methods do not help, you can simply replace zscale with scale.

  • cmd.use_zscale == True means zscale is used
  • cmd.use_zscale == False means scale is used
from vtcff import FfmpegCommand, Scale

cmd = FfmpegCommand()

assert cmd.use_zscale == True  # by default, it's 'zscale' (zimg) 
cmd.use_zscale = False  # switching to 'scale' (libswscale)

# properties affected:
cmd.scale = Scale(1920, 1080)
cmd.src_color_space = 'bt709'
cmd.dst_color_space = 'bt2020ncl'
cmd.src_range_full = True
cmd.dst_range_full = False

use_zscale=True, means that zimg will be used for conversions explicitly set by object properties. This is good because these conversions will be of high quality.

However, implicit conversions may also be required. For example, before processing 16-bit PNG with zscale, we need to convert the pixel format from rgba64be to gbrap16le 🤪. Ffmpeg will do it automatically with libswscale regardless of the use_zscale property.

Crop and scale

from vtcff import FfmpegCommand, Scale, Crop

# crop 10 pixels, then scale
a = FfmpegCommand()
a.crop = Crop(left=10)
a.scale = Scale(1920, 1080)

# scale, then crop 10 pixels
b = FfmpegCommand()
b.scale = Scale(1920, 1080)
b.crop = Crop(left=10)

Scale proportionally

from vtcff import FfmpegCommand, Scale

cmd = FfmpegCommand()

# set height to 1080, automatically compute width 
cmd.scale = Scale(-1, 1080)

# set height to 1080, select the width as the closest factor 
# of two to the proportional size 
cmd.scale = Scale(-2, 1080)

Change color range

from vtcff import FfmpegCommand

cmd = FfmpegCommand()

# Full/Data/PC range to Limited/Video/TV
cmd.src_range_full = True
cmd.dst_range_full = False

# rec.709 to rec.2020 
cmd.src_color_space = 'bt709'
cmd.dst_color_space = 'bt2020ncl'

Target formats

Encoding to Apple ProRes

from vtcff import FfmpegCommand, Prores, ProresProfile

cmd = FfmpegCommand()

# by default it will encode to ProRes 4:2:2
cmd.dst_codec_video = Prores()

# encode to ProRes 4:2:2 HQ instead 
cmd.dst_codec_video = Prores(profile=ProresProfile.HQ)

Encoding to HEVC (H.265)

from vtcff import FfmpegCommand, Hevc, VcPreset

cmd = FfmpegCommand()

# ideal quality
cmd.dst_codec_video = Hevc(lossless=True)

# best for bitrate quality
cmd.dst_codec_video = Hevc(near_lossless=True, mbps=100)

# default quality
cmd.dst_codec_video = Hevc(mbps=100)

# all modes can be tweaked with optional speed presets:
cmd.dst_codec_video = Hevc(mbps=100,
                           preset=VcPreset.N7_SLOW)

By default, the near_lossless is set to slowest possible VcPreset.N10_PLACEBO, because we are trying to maximize quality. You may want to choose a faster preset so that the result appears within a lifetime.

By default, the lossless is set to fastest possible VcPreset.N1_ULTRAFAST, because we are not losing any quality here. The resulting size will be roughly comparable to ProRes HQ/XQ and the encoding time is reasonable.

Copying streams

The media streams can be copied without re-encoding and without quality loss.

However, there may be some loss of metadata – for example, information about color ranges and color spaces.

from vtcff import FfmpegCommand, VideoCopy, NoAudio

cmd = FfmpegCommand()

# changing container from mp4 to mov
cmd.src_file = "source.mp4"
cmd.dst_file = "source.mov"

# keeping video, removing audio
cmd.dst_codec_video = VideoCopy()
cmd.dst_codec_audio = NoAudio()

Images to videos

Converting timelapses or CGI frame sequences to ProRes video file.

import subprocess
from vtcff import FfmpegCommand, Prores, ProresProfile

cmd = FfmpegCommand()

# input directory will be automatically transformed 
# to a pattern like '/my/dir_with_frames/img_%04.png'   
cmd.src_file = '/my/dir_with_frames'
cmd.src_fps = 29.97

cmd.dst_file = '/videos/timelapse.mov'
cmd.dst_codec_video = Prores(profile=ProresProfile.HQ)

# images usually have Full/Data/PC color range,
# but most NLEs assume that videos have Limited/Video/TV range
cmd.src_range_full = True
cmd.dst_range_full = False

# we will treat sRGB like Rec.709, 
# although it's a little sad
cmd.src_color_space = 'bt709'
cmd.dst_color_space = 'bt709'

# run command
subprocess.check_call(list(cmd))

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

vtcff-0.3.1.tar.gz (25.8 kB view details)

Uploaded Source

Built Distribution

vtcff-0.3.1-py3-none-any.whl (31.6 kB view details)

Uploaded Python 3

File details

Details for the file vtcff-0.3.1.tar.gz.

File metadata

  • Download URL: vtcff-0.3.1.tar.gz
  • Upload date:
  • Size: 25.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.5.0 importlib_metadata/4.8.1 pkginfo/1.7.1 requests/2.26.0 requests-toolbelt/0.9.1 tqdm/4.62.3 CPython/3.10.0

File hashes

Hashes for vtcff-0.3.1.tar.gz
Algorithm Hash digest
SHA256 c80f534ea7531cfb82e2406db800645071a55d28a91bd3eadbe597027f925134
MD5 8b361408a0b1ec8fb019f40f1fa01d55
BLAKE2b-256 2775b4cba1bce865dd05e652782ab73965bfd2de85e23000ff1870f168b8f5f0

See more details on using hashes here.

File details

Details for the file vtcff-0.3.1-py3-none-any.whl.

File metadata

  • Download URL: vtcff-0.3.1-py3-none-any.whl
  • Upload date:
  • Size: 31.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.5.0 importlib_metadata/4.8.1 pkginfo/1.7.1 requests/2.26.0 requests-toolbelt/0.9.1 tqdm/4.62.3 CPython/3.10.0

File hashes

Hashes for vtcff-0.3.1-py3-none-any.whl
Algorithm Hash digest
SHA256 f59445fff15f549fb8564f11a83b2d3d34326175d0e96a359aa683610af7671a
MD5 7a73640dc44be5b237841bcaa4df7b3a
BLAKE2b-256 f9d55eb07f1131f772d6db52c0511162f8700f428eae9129109343de00a9e7d2

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page