Shortcuts

ffmpeg_utils

Module to change the configuration of FFmpeg libraries (such as libavformat).

It affects functionalities in torchaudio.io (and indirectly torchaudio.load()).

clear_cuda_context_cache

torchaudio.utils.ffmpeg_utils.clear_cuda_context_cache()[source]

Clear the CUDA context used by CUDA Hardware accelerated video decoding

get_audio_decoders

torchaudio.utils.ffmpeg_utils.get_audio_decoders() Dict[str, str][source]

Get the available audio decoders.

Returns:

Mapping from decoder short name to long name.

Return type:

Dict[str, str]

Example
>>> for k, v in get_audio_decoders().items():
>>>     print(f"{k}: {v}")
... a64: a64 - video for Commodore 64
... ac3: raw AC-3
... adts: ADTS AAC (Advanced Audio Coding)
... adx: CRI ADX
... aiff: Audio IFF

get_audio_encoders

torchaudio.utils.ffmpeg_utils.get_audio_encoders() Dict[str, str][source]

Get the available audio encoders.

Returns:

Mapping from encoder short name to long name.

Return type:

Dict[str, str]

Example
>>> for k, v in get_audio_encoders().items():
>>>     print(f"{k}: {v}")
... comfortnoise: RFC 3389 comfort noise generator
... s302m: SMPTE 302M
... aac: AAC (Advanced Audio Coding)
... ac3: ATSC A/52A (AC-3)
... ac3_fixed: ATSC A/52A (AC-3)
... alac: ALAC (Apple Lossless Audio Codec)

get_build_config

torchaudio.utils.ffmpeg_utils.get_build_config() str[source]

Get the FFmpeg build configuration

Returns:

Build configuration string.

Return type:

str

Example
>>> print(get_build_config())
--prefix=/Users/runner/miniforge3 --cc=arm64-apple-darwin20.0.0-clang --enable-gpl --enable-hardcoded-tables --enable-libfreetype --enable-libopenh264 --enable-neon --enable-libx264 --enable-libx265 --enable-libaom --enable-libsvtav1 --enable-libxml2 --enable-libvpx --enable-pic --enable-pthreads --enable-shared --disable-static --enable-version3 --enable-zlib --enable-libmp3lame --pkg-config=/Users/runner/miniforge3/conda-bld/ffmpeg_1646229390493/_build_env/bin/pkg-config --enable-cross-compile --arch=arm64 --target-os=darwin --cross-prefix=arm64-apple-darwin20.0.0- --host-cc=/Users/runner/miniforge3/conda-bld/ffmpeg_1646229390493/_build_env/bin/x86_64-apple-darwin13.4.0-clang  # noqa

get_demuxers

torchaudio.utils.ffmpeg_utils.get_demuxers() Dict[str, str][source]

Get the available demuxers.

Returns:

Mapping from demuxer (format) short name to long name.

Return type:

Dict[str, str]

Example
>>> for k, v in get_demuxers().items():
>>>     print(f"{k}: {v}")
... aa: Audible AA format files
... aac: raw ADTS AAC (Advanced Audio Coding)
... aax: CRI AAX
... ac3: raw AC-3

get_input_devices

torchaudio.utils.ffmpeg_utils.get_input_devices() Dict[str, str][source]

Get the available input devices.

Returns:

Mapping from device short name to long name.

Return type:

Dict[str, str]

Example
>>> for k, v in get_input_devices().items():
>>>     print(f"{k}: {v}")
... avfoundation: AVFoundation input device
... lavfi: Libavfilter virtual input device

get_input_protocols

torchaudio.utils.ffmpeg_utils.get_input_protocols() List[str][source]

Get the supported input protocols.

Returns:

The names of supported input protocols

Return type:

List[str]

Example
>>> print(get_input_protocols())
... ['file', 'ftp', 'hls', 'http','https', 'pipe', 'rtmp', 'tcp', 'tls', 'udp', 'unix']

get_log_level

torchaudio.utils.ffmpeg_utils.get_log_level() int[source]

Get the log level of FFmpeg.

See set_log_level() for the detailo.

get_muxers

torchaudio.utils.ffmpeg_utils.get_muxers() Dict[str, str][source]

Get the available muxers.

Returns:

Mapping from muxer (format) short name to long name.

Return type:

Dict[str, str]

Example
>>> for k, v in get_muxers().items():
>>>     print(f"{k}: {v}")
... a64: a64 - video for Commodore 64
... ac3: raw AC-3
... adts: ADTS AAC (Advanced Audio Coding)
... adx: CRI ADX
... aiff: Audio IFF

get_output_devices

torchaudio.utils.ffmpeg_utils.get_output_devices() Dict[str, str][source]

Get the available output devices.

Returns:

Mapping from device short name to long name.

Return type:

Dict[str, str]

Example
>>> for k, v in get_output_devices().items():
>>>     print(f"{k}: {v}")
... audiotoolbox: AudioToolbox output device

get_output_protocols

torchaudio.utils.ffmpeg_utils.get_output_protocols() List[str][source]

Get the supported output protocols.

Returns:

The names of supported output protocols

Return type:

list of str

Example
>>> print(get_output_protocols())
... ['file', 'ftp', 'http', 'https', 'md5', 'pipe', 'prompeg', 'rtmp', 'tee', 'tcp', 'tls', 'udp', 'unix']

get_versions

torchaudio.utils.ffmpeg_utils.get_versions() Dict[str, Tuple[int]][source]

Get the versions of FFmpeg libraries

Returns:

mapping from library names to version string,

i.e. “libavutil”: (56, 22, 100).

Return type:

dict

Tutorials using get_versions:
AudioEffector Usages

AudioEffector Usages

AudioEffector Usages
Accelerated video encoding with NVENC

Accelerated video encoding with NVENC

Accelerated video encoding with NVENC
StreamWriter Basic Usage

StreamWriter Basic Usage

StreamWriter Basic Usage
Accelerated video decoding with NVDEC

Accelerated video decoding with NVDEC

Accelerated video decoding with NVDEC

get_video_decoders

torchaudio.utils.ffmpeg_utils.get_video_decoders() Dict[str, str][source]

Get the available video decoders.

Returns:

Mapping from decoder short name to long name.

Return type:

Dict[str, str]

Example
>>> for k, v in get_video_decoders().items():
>>>     print(f"{k}: {v}")
... aasc: Autodesk RLE
... aic: Apple Intermediate Codec
... alias_pix: Alias/Wavefront PIX image
... agm: Amuse Graphics Movie
... amv: AMV Video
... anm: Deluxe Paint Animation
Tutorials using get_video_decoders:
Accelerated video decoding with NVDEC

Accelerated video decoding with NVDEC

Accelerated video decoding with NVDEC

get_video_encoders

torchaudio.utils.ffmpeg_utils.get_video_encoders() Dict[str, str][source]

Get the available video encoders.

Returns:

Mapping from encoder short name to long name.

Return type:

Dict[str, str]

Example
>>> for k, v in get_audio_encoders().items():
>>>     print(f"{k}: {v}")
... a64multi: Multicolor charset for Commodore 64
... a64multi5: Multicolor charset for Commodore 64, extended with 5th color (colram)
... alias_pix: Alias/Wavefront PIX image
... amv: AMV Video
... apng: APNG (Animated Portable Network Graphics) image
... asv1: ASUS V1
... asv2: ASUS V2
Tutorials using get_video_encoders:
Accelerated video encoding with NVENC

Accelerated video encoding with NVENC

Accelerated video encoding with NVENC

set_log_level

torchaudio.utils.ffmpeg_utils.set_log_level(level: int)[source]

Set the log level of FFmpeg (libavformat etc)

Parameters:

level (int) –

Log level. The larger, the more verbose.

The following values are common values, the corresponding ffmpeg’s -loglevel option value and desription.

  • -8 (quiet): Print no output.

  • 0 (panic): Something went really wrong and we will crash now.

  • 8 (fatal): Something went wrong and recovery is not possible. For example, no header was found for a format which depends on headers or an illegal combination of parameters is used.

  • 16 (error): Something went wrong and cannot losslessly be recovered. However, not all future data is affected.

  • 24 (warning): Something somehow does not look correct. This may or may not lead to problems.

  • 32 (info): Standard information.

  • 40 (verbose): Detailed information.

  • 48 (debug): Stuff which is only useful for libav* developers.

  • 56 (trace): Extremely verbose debugging, useful for libav* development.

Tutorials using set_log_level:
Accelerated video encoding with NVENC

Accelerated video encoding with NVENC

Accelerated video encoding with NVENC

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources