video/example-add-captions708
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

Use google cloud speech-to-text plugin

Browse source
This commit is contained in:
Rafael Caricio 2022-04-12 11:05:16 +02:00
parent b1766f2b0a
commit e22a0107b3
Signed by: rafaelcaricio
GPG key ID: 3C86DBCE8E93C947
1 changed files with 20 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

31
src/main.rs
View file

@ -13,13 +13,19 @@ fn main() -> eyre::Result<()> {
gstrsclosedcaption::plugin_register_static()?; gstrsclosedcaption::plugin_register_static()?;
gstvosk::plugin_register_static()?; gstvosk::plugin_register_static()?;
// ristsrc name=rist_src address="0.0.0.0" ! rtpmp2tdepay name=rtmpdepay ! decodebin name=multiplexer
// uridecodebin name=demuxer uri=file:///Users/rafael.caricio/video.mkv
//
// demuxer. ! videorate ! video/x-raw,framerate=(fraction)30/1 ! ccextractor remove-caption-meta=true ! transcriberbin name=trans latency=30000
// demuxer. ! audio/x-raw ! audiorate ! audioconvert ! audioresample ! trans.sink_audio
let pipeline = gst::parse_launch( let pipeline = gst::parse_launch(
r#" r#"
uridecodebin name=demuxer uri=file:///Users/rafael.caricio/video.mkv uridecodebin name=multiplexer uri=file:///Users/rafael.caricio/video.mkv
demuxer. ! videorate ! video/x-raw,framerate=(fraction)30/1 ! transcriberbin name=trans latency=30000 multiplexer. ! videorate ! video/x-raw,framerate=(fraction)30/1 ! ccextractor remove-caption-meta=true ! trans.sink_video
demuxer. ! audio/x-raw ! audiorate ! audioconvert ! audioresample ! trans.sink_audio multiplexer. ! audio/x-raw ! audiorate ! audioconvert ! audioresample ! transcriberbin name=trans
trans.src_video ! cea608overlay black-background=1 ! autovideosink trans.src_video ! cea608overlay black-background=1 ! autovideosink
trans.src_audio ! autoaudiosink trans.src_audio ! autoaudiosink
@ -31,17 +37,20 @@ fn main() -> eyre::Result<()> {
info!("Starting pipeline..."); info!("Starting pipeline...");
let demuxer = pipeline.by_name("demuxer").unwrap(); // let demuxer = pipeline.by_name("demuxer").unwrap();
demuxer.connect_pad_added(|_, pad| { // demuxer.connect_pad_added(|_, pad| {
let name = pad.name(); // let name = pad.name();
let caps = pad.caps().unwrap(); // let caps = pad.caps().unwrap();
let caps_type = caps.structure(0).unwrap().name(); // let caps_type = caps.structure(0).unwrap().name();
info!("Pad {} added with caps {}", name, caps_type); // info!("Pad {} added with caps {}", name, caps_type);
}); // });
let transcriber = gst::ElementFactory::make("vosk_transcriber", None).expect("Could not instantiate Vosk transcriber"); let transcriber = gst::ElementFactory::make("gspeechtotext", None).expect("Could not instantiate Google transcriber");
transcriber.set_property("auth-json-file",
"/Users/rafael.caricio/development/live/google-cloud-playground/i-centralvideo-dictate-dev-c184dd68967a.json");
let transcriber_bin = pipeline.by_name("trans").expect("Trans bin"); let transcriber_bin = pipeline.by_name("trans").expect("Trans bin");
transcriber_bin.set_property("transcriber", transcriber); transcriber_bin.set_property("transcriber", transcriber);
transcriber_bin.set_property("latency", 45_000_u32);
let context = glib::MainContext::default(); let context = glib::MainContext::default();