Docker image with ffmpeg and ffprobe built as hardened static PIE binaries with no external dependencies that can be used with any base image.
See Dockerfile for versions used. In general, master should have the latest stable version of ffmpeg and below libraries. Versions are kept up to date automatically using bump.
Use mwader/static-ffmpeg
from Docker Hub or build the image yourself.
In Dockerfile
COPY --from=mwader/static-ffmpeg:7.0.1 /ffmpeg /usr/local/bin/
COPY --from=mwader/static-ffmpeg:7.0.1 /ffprobe /usr/local/bin/
Run directly
docker run -i --rm -u $UID:$GROUPS -v "$PWD:$PWD" -w "$PWD" mwader/static-ffmpeg:7.0.1 -i file.wav file.mp3
docker run -i --rm -u $UID:$GROUPS -v "$PWD:$PWD" -w "$PWD" --entrypoint=/ffprobe mwader/static-ffmpeg:7.0.1 -i file.wav
As shell alias
alias ffmpeg='docker run -i --rm -u $UID:$GROUPS -v "$PWD:$PWD" -w "$PWD" mwader/static-ffmpeg:7.0.1'
alias ffprobe='docker run -i --rm -u $UID:$GROUPS -v "$PWD:$PWD" -w "$PWD" --entrypoint=/ffprobe mwader/static-ffmpeg:7.0.1'
- fontconfig
- gray (full grayscale support)
- iconv (from musl)
- lcms2
- libaom
- libaribb24
- libass
- libbluray
- libdav1d
- libdavs2
- libfdk-aac (only if explicitly enabled during build, see below)
- libfreetype
- libfribidi
- libgme
- libgsm
- libharfbuzz
- libjxl
- libkvazaar
- libmodplug
- libmp3lame
- libmysofa
- libopencore
- libopenjpeg
- libopus
- librabbitmq
- librav1e
- librsvg
- librtmp
- librubberband
- libshine
- libsnappy
- libsoxr
- libspeex
- libsrt
- libssh
- libsvtav1
- libtheora
- libtwolame
- libuavs3d
- libvidstab
- libvmaf
- libvo-amrwbenc
- libvorbis
- libvpx
- libwebp
- libx264
- libx265 (multilib with support for 10 and 12 bits)
- libxavs2
- libxevd
- libxeve
- libxml2
- libxvid
- libzimg
- openssl
- and all native ffmpeg codecs, formats, filters etc.
/ffmpeg
ffmpeg binary/ffprobe
ffprobe binary/doc
Documentation/versions.json
JSON file with build versions of ffmpeg and libraries./etc/ssl/cert.pem
CA certs to make-tls_verify 1 -ca_file /etc/ssl/cert.pem
work if running image directly- Fonts, fontconfig config and pre-populated cache:
/etc/fonts
/usr/share/fonts
/usr/share/consolefonts
/var/cache/fontconfig
latest
Latest master build.
MAJOR.MINOR.PATCH[-BUILD]
Specific version of FFmpeg with the features that was in master at the time of tagging.
-BUILD
means that was an additional build with that version to add of fix something.
Binaries are built with various hardening features but it's still a good idea to run them as non-root even when used inside a container, especially so if running on input files that you don't control.
Due to license issues the docker image does not include libfdk-aac by default. A docker image including libfdk-aac can be built by passing a non empty value to the build-arg ENABLE_FDKAAC
, example below.
docker build --build-arg ENABLE_FDKAAC=1 . -t my-ffmpeg-static:latest
The image ships with some basic fonts (font-terminus font-inconsolata font-dejavu font-awesome
) that can be used when running the image directly. If your copying the binaries into some image you have to install fonts somehow. How to do this depends a bit on distributions but in general look for font packages and how to make fontconfig know about them.
- Alpine Linux see https://wiki.alpinelinux.org/wiki/Fonts
- Debian/Ubuntu see https://wiki.debian.org/Fonts
Let's say you're building a docker image for your application that requires the ffmpeg
and ffprobe
binaries, and want to have a dedicated directory for storing fonts, e.g. /app/fonts
. In order to achieve this, let's take a look at the following steps.
- A custom config file containing the fonts directory you want to use has to be added to the image. Here, a file
50-custom.conf
will be added to the/etc/fonts/conf.d
directory, containing an additional/app/fonts
directory for Fontconfig to use.
COPY <<EOF /etc/fonts/conf.d/50-custom.conf
<?xml version="1.0"?>
<!DOCTYPE fontconfig SYSTEM "urn:fontconfig:fonts.dtd">
<fontconfig>
<dir>/app/fonts</dir>
</fontconfig>
EOF
Tip
Check the Fontconfig User Documentation for examples and available options for your custom fontconfig file.
- Copy the
ffmpeg
(andffprobe
) binaries.
COPY --from=mwader/static-ffmpeg:7.0.1 /ffmpeg /usr/bin/
COPY --from=mwader/static-ffmpeg:7.0.1 /ffprobe /usr/bin/
- Make sure the
/app/fonts
directory exist in your image.
WORKDIR /app
RUN <<EOT bash
set -ex
mkdir -p ./fonts
EOT
If you followed the steps above, your image should look something like this.
FROM python:3.11-slim-bookworm AS base
COPY <<EOF /etc/fonts/conf.d/50-custom.conf
<?xml version="1.0"?>
<!DOCTYPE fontconfig SYSTEM "urn:fontconfig:fonts.dtd">
<fontconfig>
<dir>/app/fonts</dir>
</fontconfig>
EOF
FROM base as ffmpeg
COPY --from=mwader/static-ffmpeg:7.0.1 /ffmpeg /usr/bin/
COPY --from=mwader/static-ffmpeg:7.0.1 /ffprobe /usr/bin/
FROM ffmpeg AS app
WORKDIR /app
RUN <<EOT bash
set -ex
mkdir -p ./fonts
EOT
You can now build the image and run a container that has volume mounted to /app/fonts
.
docker build -t ffmpeg-fonts-image .
docker run -i --rm \
-u "$UID:$GROUPS" \
-v "$PWD:$PWD" \
-v "$PWD/fonts:/app/fonts" \
-v "$PWD/cache:/var/cache/fontconfig" \
-w "$PWD" \
ffmpeg-fonts-image \
ffmpeg -v debug -y -f lavfi -i 'color=white,drawtext=text=Test:fontfile=Arial' -t 1s /app/output.mp4
Inspecting the log, you will find that the font is located in /app/fonts
.
Opening an input file: color=white,drawtext=text=Test:fontfile=Arial.
[AVFilterGraph @ 0x7fb481827800] Setting 'color' to value 'white'
[AVFilterGraph @ 0x7fb481827800] Setting 'text' to value 'Test'
[AVFilterGraph @ 0x7fb481827800] Setting 'fontfile' to value 'Arial'
[Parsed_drawtext_1 @ 0x7fb481827cc0] Using "/app/fonts/arial.ttf"
Note
The volume mount for /var/cache/fontconfig
is optional. If you run the container as a non-root user, just make sure the cache directory exists on host before mounting, else it will still show the debug message Fontconfig error: No writable cache directories
. This error does not further impact Fontconfig in locating the appropriate fonts.
Binaries are built with TLS support but, by default, ffmpeg currently do
not do certificate verification. To enable verification you need to run
ffmpeg with -tls_verify 1
and -ca_file /path/to/cert.pem
.
- Alpine Linux at
/etc/ssl/cert.pem
- Debian/Ubuntu install the
ca-certificates
package at it will be available at/etc/ssl/certs/ca-certificates.crt
.
Since version 5.0.1-3 dockerhub images are multi-arch amd64 and arm64 images.
This will copy ffmpeg
and ffprobe
to the current directory:
docker run --rm -v "$PWD:/out" $(echo -e 'FROM alpine\nCOPY --from=mwader/static-ffmpeg:7.0.1 /ff* /\nENTRYPOINT cp /ff* /out' | docker build -q -)
docker run --rm mwader/static-ffmpeg -v quiet -f data -i versions.json -map 0 -c copy -f data -
This could happen if the hostname resolve to more IP-addresses than can fit in DNS UDP packet (probably 512 bytes) causing the response to be truncated. Usually clients should then switch to TCP and redo the query. This should only be a problem with version 6.0-1 or earlier of this image that uses musl libc 1.2.3 or older.
This could be because the statically linked openssl version in the static-ffmpeg binaries are not compatible with the (possibly distro modified openssl) configuration files found in the filesystem. The error is about openssl encountering an option that it does not know about.
Possible workarounds:
- Add
config_diagnostics = 0
to the openssl config to ignore unknown options with the risk of ignoring real problems. - Use ffmpeg option
-reconnect_on_network_error true
to ignore the error but will still warn.
See these references for further discussion and workarounds:
- First call to SSL_CTX_new is failing on AL2023 (3.0.12)
- OpenSSL issue with binary outside container (RedHat/Fedora specific)
Multi-arch dockerhub images are built using pyldin601/build-multiarch-on-aws-spots. See build-multiarch.yml for config. Thanks to @pyldin601 for making this possible.
Feel free to create issues or PRs if you have any improvements or encounter any problems. Please also consider making a donation to the FFmpeg project or to other projects used by this image if you find it useful.
Please also be mindful of the license limitations used by libraries this project uses and your own usage and potential distribution of such.
- Add libplacebo, chromaprint, etc. ...
- Add vvenc/vvdec support once in stable
- Add acceleration support (GPU, CUDA, ...)
- Add *.a *.so libraries, headers and pkg-config somehow