Revert "Upgrade to FFmpeg5 (#3298)" (#3377)

Summary:
This reverts commit d38a7854.

This is temporary revert to unblock unit test migration from circleci to github

Pull Request resolved: https://github.com/pytorch/audio/pull/3377

Reviewed By: mthrok

Differential Revision: D46230498

Pulled By: atalman

fbshipit-source-id: 000d8a9ca00750fc1ca61f4c2cdd6e930a5ce46d

.circleci/unittest/linux/scripts/setup_env.sh

@@ -36,4 +36,4 @@ conda activate "${env_dir}"
 
 # 3. Install minimal build tools
 pip --quiet install cmake ninja
-conda install --quiet -y -c conda-forge 'ffmpeg==5.1' pkg-config
+conda install --quiet -y 'ffmpeg>=4.1' pkg-config

.circleci/unittest/windows/scripts/setup_env.sh

@@ -35,4 +35,4 @@ conda activate "${env_dir}"
 
 # 3. Install minimal build tools
 pip --quiet install cmake ninja
-conda install --quiet -y -c conda-forge 'ffmpeg==5.1'
+conda install --quiet -y 'ffmpeg>=4.1'

.github/workflows/build_docs.yml

@@ -106,6 +106,8 @@ jobs:
           "${CUDATOOLKIT}"
 
         # Install torchaudio
+        # TODO: Enable NVDec/NVEnc
+        conda install --quiet -y 'ffmpeg>=4.1' pkg-config
         pip --quiet install cmake>=3.18.0 ninja
         cd packaging
         . ./pkg_helpers.bash

.github/workflows/unittest-linux-gpu.yml

@@ -55,7 +55,7 @@ jobs:
           "${CUDATOOLKIT}"
 
         # Install torchaudio
-        conda install --quiet -y -c conda-forge 'ffmpeg==5.1' pkg-config
+        conda install --quiet -y 'ffmpeg>=4.1' pkg-config
         python3 -m pip --quiet install cmake>=3.18.0 ninja
         USE_FFMPEG=1 python3 -m pip install -v -e . --no-use-pep517
 

docs/source/installation.rst

@@ -16,7 +16,7 @@ Please refer to https://pytorch.org/get-started/locally/ for the details.
    each of which requires a corresponding PyTorch distribution.
 
 .. note::
-   This software was compiled against an unmodified copy of FFmpeg (licensed under `the LGPLv2.1 <https://github.com/FFmpeg/FFmpeg/blob/0e15444aceca0e78f99f3d67758eb79d11b86599/COPYING.LGPLv2.1>`_), with the specific rpath removed so as to enable the use of system libraries. The LGPL source can be downloaded `here <https://github.com/FFmpeg/FFmpeg/releases/tag/n5.0.3>`_.
+   This software was compiled against an unmodified copy of FFmpeg (licensed under `the LGPLv2.1 <https://github.com/FFmpeg/FFmpeg/blob/a5d2008e2a2360d351798e9abe883d603e231442/COPYING.LGPLv2.1>`_), with the specific rpath removed so as to enable the use of system libraries. The LGPL source can be downloaded `here <https://github.com/FFmpeg/FFmpeg/releases/tag/n4.1.8>`_.
 
 Dependencies
 ------------
@@ -31,8 +31,8 @@ Optional Dependencies
 * `FFmpeg <https://ffmpeg.org>`_.
 
   Required to use :py:mod:`torchaudio.io` module.
-  TorchAudio official binary distributions are compatible with FFmpeg 5.
-  If you need to use FFmpeg 6, please build TorchAudio from source.
+  TorchAudio official binary distributions are compatible with FFmpeg 4.1 to 4.4.
+  If you need to use FFmpeg 5, please build TorchAudio from source.
 
 * `sentencepiece <https://pypi.org/project/sentencepiece/>`_
 

examples/tutorials/device_asr.py

@@ -10,11 +10,11 @@ on laptop.
 
 .. note::
 
-   This tutorial requires FFmpeg libraries (>=5.0, <6) and SentencePiece.
+   This tutorial requires FFmpeg libraries (>=4.1, <4.4) and SentencePiece.
 
    There are multiple ways to install FFmpeg libraries.
    If you are using Anaconda Python distribution,
-   ``conda install -c conda-forge 'ffmpeg<6'`` will install
+   ``conda install 'ffmpeg<4.4'`` will install
    the required FFmpeg libraries.
 
    You can install SentencePiece by running ``pip install sentencepiece``.

examples/tutorials/effector_tutorial.py

@@ -13,11 +13,11 @@ apply various effects and codecs to waveform tensor.
 #
 # .. note::
 #
-#    This tutorial requires FFmpeg libraries (>=5.0, <6).
+#    This tutorial requires FFmpeg libraries (>=4.1, <5).
 #
 #    There are multiple ways to install FFmpeg libraries.
 #    If you are using Anaconda Python distribution,
-#    ``conda install -c conda-forge 'ffmpeg<6'`` will install
+#    ``conda install -c anaconda 'ffmpeg<5'`` will install
 #    the required libraries.
 #
 

examples/tutorials/online_asr_tutorial.py

@@ -13,11 +13,11 @@ to perform online speech recognition.
 #
 # .. note::
 #
-#    This tutorial requires FFmpeg libraries (>=5, <6) and SentencePiece.
+#    This tutorial requires FFmpeg libraries (>=4.1, <4.4) and SentencePiece.
 #
 #    There are multiple ways to install FFmpeg libraries.
 #    If you are using Anaconda Python distribution,
-#    ``conda install -c conda-forge 'ffmpeg<6'`` will install
+#    ``conda install 'ffmpeg<4.4'`` will install
 #    the required FFmpeg libraries.
 #
 #    You can install SentencePiece by running ``pip install sentencepiece``.

examples/tutorials/streamreader_basic_tutorial.py

@@ -14,11 +14,11 @@ libavfilter provides.
 #
 # .. note::
 #
-#    This tutorial requires FFmpeg libraries (>=5.0, <6).
+#    This tutorial requires FFmpeg libraries (>=4.1, <4.4).
 #
 #    There are multiple ways to install FFmpeg libraries.
 #    If you are using Anaconda Python distribution,
-#    ``conda install -c conda-forge 'ffmpeg<6'`` will install
+#    ``conda install -c anaconda 'ffmpeg<4.4'`` will install
 #    the required libraries.
 #
 

examples/tutorials/streamwriter_advanced.py

@@ -23,14 +23,17 @@ play audio and video.
 #
 # .. note::
 #
-#    This tutorial requires FFmpeg libraries (>=5.0, <6).
+#    This tutorial requires torchaudio nightly build and FFmpeg libraries (>=4.1, <4.4).
+#
+#    To install torchaudio nightly build, please refer to
+#    https://pytorch.org/get-started/locally/ .
+#
 #
 #    There are multiple ways to install FFmpeg libraries.
 #    If you are using Anaconda Python distribution,
-#    ``conda install -c conda-forge 'ffmpeg<6'`` will install
-#    the required libraries.
-#    This distribution, however, does not have SDL plugin, so
-#    it cannot play video.
+#    ``conda install 'ffmpeg<4.4'`` will install the required FFmpeg libraries,
+#    however, this distribution does not have SDL plugin, so it cannot play
+#    video.
 #
 
 ######################################################################

examples/tutorials/streamwriter_basic_tutorial.py

@@ -13,12 +13,14 @@ encode and save audio/video data into various formats/destinations.
 #
 # .. note::
 #
-#    This tutorial requires FFmpeg libraries (>=5.0, <6).
+#    This tutorial requires torchaudio nightly build and FFmpeg libraries (>=4.1, <4.4).
+#
+#    To install torchaudio nightly build, please refer to
+#    https://pytorch.org/get-started/locally/ .
 #
 #    There are multiple ways to install FFmpeg libraries.
 #    If you are using Anaconda Python distribution,
-#    ``conda install -c conda-forge 'ffmpeg<6'`` will install
-#    the required libraries.
+#    ``conda install 'ffmpeg<4.4'`` will install the required FFmpeg libraries.
 #
 
 ######################################################################

packaging/ffmpeg/build.sh

@@ -32,7 +32,7 @@ cd "${build_dir}"
 # NOTE:
 # When changing the version of FFmpeg, update the README so that the link to the source points
 # the same version.
-curl -LsS -o ffmpeg.tar.gz https://github.com/FFmpeg/FFmpeg/archive/refs/tags/n5.0.3.tar.gz
+curl -LsS -o ffmpeg.tar.gz https://github.com/FFmpeg/FFmpeg/archive/refs/tags/n4.1.8.tar.gz
 tar -xf ffmpeg.tar.gz --strip-components 1
 ./configure \
     --prefix="${prefix}" \
@@ -72,11 +72,11 @@ ls ${prefix}/*
 # macOS: Fix rpath so that the libraries are searched dynamically in user environment.
 # In Linux, this is handled by `--enable-rpath` flag.
 if [[ "$(uname)" == Darwin ]]; then
-    avcodec=libavcodec.59
-    avdevice=libavdevice.59
-    avfilter=libavfilter.8
-    avformat=libavformat.59
-    avutil=libavutil.57
+    avcodec=libavcodec.58
+    avdevice=libavdevice.58
+    avfilter=libavfilter.7
+    avformat=libavformat.58
+    avutil=libavutil.56
 
     otool="/usr/bin/otool"
     # NOTE: miniconda has a version of otool and install_name_tool installed and we want