Handle post-production in an automated fashion. Joining video track, slides track, and sound track(s) is done by using file dates to sync the various sources. All you need to do is specify the start and end time of the main video. Ensure all recording devices are time-synced, for instance, by connecting them to the same Wi-Fi network.
poetry run python src/produce_video_v2.py ht template/ht/talk1/config.yml 7 --no-intro
poetry run python src/produce_video_v2.py ht template/ht/talk1/config.yml
poetry run python src/fetch_talk_info.py https://humantalks.com/cities/grenoble/events/609
poetry run python src/make_thumbnail.py ht template/ht/talk1/config.yml- Transition-fade between intro and main video
- When capturing, you can start any time; all you have to provide is the start time of one video
- Automatic adjustment of tracks based on media date metadata
extra_offsetcan be used to compensate for bad dates
Either provide a ready-made intro file or generate one automatically with:
- Presentation title
- Speaker name
- Background
- Logo
- Jingle
python3 src/produce_video_v2.py [conference_theme] session-config.yaml [max time seconds] <--no-intro>conference_themeis eitherhtoralpescraft. It guides the background and the intro.max_time_secondsis used when you want to generate a preview and avoid too much processing time.
The timing is never completely right. To calibrate, the current best solution is to generate a 7-second long video, check the result, and then adapt the extra_offset to the sound file. For example:
python3 src/produce_video_v2.py ht session-config.yaml 7 --no-introconfig.yml is a file of the form (in its simplest form):
title: "awesome session"
speaker_name: "John Doe"
speaker:
parts:
# This means that the video will start at 1 minute and 8 seconds from the start of the speaker video.
# Sound and slides will be synced accordingly.
- start: "1:08"
stop: "55:15"
sound:
slides:It uses conventions for the file names and locations. Necessary directory layout:
.
└── to-process
├── jingle.mp4 # Jingle video file for the intro (if dynamically generated)
├── logo.webp # Logo image for the conference theme
├── talk1 # Directory for the first talk/session
│ ├── config.yml # Configuration file for the video processing
│ ├── slides.mkv # Video file of the slides for this talk
│ ├── sound.mp3 # Audio file of the talk's sound
│ ├── speaker.mp4 # Video file of the speaker for this talk
├── talk2 # Directory for the second talk/session
│ ├── config.yml # Configuration file for the video processing
│ ├── slides.mkv # Video file of the slides for this talk
│ ├── sound.mp3 # Audio file of the talk's sound
│ ├── speaker.mp4 # Video file of the speaker for this talk
├── talk3 # Directory for the third talk/session
│ ├── config.yml # Configuration file for the video processing
│ ├── slides.mkv # Video file of the slides for this talk
│ ├── sound.mp3 # Audio file of the talk's sound
│ ├── speaker.mp4 # Video file of the speaker for this talk
└── talk4 # Directory for the fourth talk/session
├── config.yml # Configuration file for the video processing
├── slides.mkv # Video file of the slides for this talk
├── sound.mp3 # Audio file of the talk's sound
├── speaker.mp4 # Video file of the speaker for this talk
logo.webp: The logo image used for the conference theme.jingle.mp4: A jingle video file used for the intro if the intro is dynamically generated.talk1,talk2,talk3,talk4: Directories for each individual talk/session. Each directory should contain:config.yml: Configuration file with the session details and synchronization settings.slides.mkv: Video file of the slides for the talk.sound.mp3: Audio file of the talk's sound.speaker.mp4: Video file of the speaker's presentation.
pip3 install poetry
poetry installThis script allows you to retrieve information from a HumanTalks event and generate YAML configuration files for each talk, as well as download the speakers' images.
poetry run python src/fetch_talk_info.py https://humantalks.com/cities/grenoble/events/609The to-proceed directory will have the following structure:
to-proceed/
├── talk1/
│ ├── speaker.webp # Speaker's image
│ └── config.yml # YAML configuration file with talk information
├── talk2/
│ ├── speaker.webp # Speaker's image
│ └── config.yml # YAML configuration file with talk information
├── talk3/
│ ├── speaker.webp # Speaker's image
│ └── config.yml # YAML configuration file with talk information
└── ...
This script allows you to retrieve information from a HumanTalks event and generate YAML configuration files for each talk, as well as download the speakers' images.
poetry run python src/make_thumbnail.py ht template/ht/talk1/config.ymlTo set up your environment using Docker Compose, follow these steps:
Make sure you have Docker and Docker Compose installed on your machine. You can download and install them from Docker.
docker-compose build
docker-compose up -d
docker-compose exec alpescraft-videos /bin/bashOnce inside the container, you can generate a test video by running:
poetry run python src/produce_video_v2.py ht template/ht/talk1/config.yml 7 --no-introTo generate the full video, run:
poetry run python src/produce_video_v2.py ht template/ht/talk1/config.yml| Command | Description |
|---|---|
docker-compose up |
Start the services defined in docker-compose.yml |
docker-compose up -d |
Start the services in detached mode |
docker-compose down |
Stop the services and clean up |
docker-compose logs |
Show logs from the services |
docker-compose exec app /bin/bash |
Access the app container in interactive mode |
docker-compose exec web /bin/bash |
Access the web container in interactive mode |
docker-compose build |
Build or rebuild the images |
docker-compose stop |
Stop the services without removing the containers |
- I used DaVinci Resolve to crop the video and flip it because the slides are on the right, and people tend to look on the left.
- Ensure all file paths and configurations are correct for successful video generation.
- Docker Compose Documentation
- Dockerfile Reference
- Docker Compose CLI Reference
- Python HTTP Server Documentation