Transcodes find-sign "search-data" format in to "encoded-search-data" format with multiple codec/resolution variants. This data can be used to directly build web interfaces like auslan-find-sign/find-sign-website. Audio tracks are stripped from videos, and clipping rules are implemented.
This follows the same format as search-data, but media is transcoded so each media entry like:
{ "method": "fetch", "url": "path/to/media.mkv", "version": "1234" }is replaced with a MediaSet:
{
"type": "video" | "audio",
"source": { "method": "fetch", "url": "path/to/media.mkv", "version": "1234" },
"thumbnail": "path/to/thumbnail.webp",
"timestamp": number,
"encodes": [
{ "type": "video/mp4; codecs=\"avc1.4d002a\"", "width": 512, "height": 288, "src": "path/to/encode.mp4" },
{ "type": "video/webm; codecs=\"vp8, vorbis\"", "width": 1024, "height": 576, "src": "path/to/encode.webm" }
]
}type is either video or audio. Find sign doesn't use audio.
source is a copy of the thing that was encoded, to preserve url and version for cache invalidation when the source changes.
timestamp is an epochMs remembering when the video was converted. This might be used in the future to re-encode media after a cache duration, to take advantage of advances in encoder technology.
encodes is a list of <source> elements that can be constructed inside a <video> or <audio> tag to play the encodes in a web browser. src may contain relative paths, so should be interpreted relative to the location of the encoded-search-data.json file.