erikstmartin/rtmp-switcher

rtmp-switcher's Issues

flush input buffers before removing?

Weird glitching video when disconnecting an input. This could be data left in the buffers?

input: configurable volume

input- ability to add RTP and UDP sources to mixer. fallbackswitch in event there is loss of stream.

Hello,
Love what you have done here with rust.. I'm new to rust and have been using gstreamer with glib python bindings.
Is there any appetite to include a UDP/RTP input to connect to the mixer.? Would be great. In addition, could the new fallbackswitch element. (written in rust) could be used to allow broken streams to fall back to a black videotestsrc and audio silence to keep the pipeline from eos?

logging

Look into github.com/slog-rs/slog

docker: Update container to use Gstreamer 1.18

Will need to compile gstreamer from source https://gitlab.freedesktop.org/gstreamer/gst-build and copy artifacts over to runtime container.

gstreamer: Pipeline breaks when using different output types

If we try to output to both autosink and rtmp the pipeline never leaves the Paused state.

input: Image input type

mixer: Transitions

Adding a video after the mixer has been playing causes video to fast-forward

This has something to do with the internal clock. The video tries to catch up to the internal clock, which causes it to fast-forward. We don't want this to happen.

output: Add ability to change resolution and volume

mixer: Add lower-thirds

mixer: Test inputs of different sizes.

Does the current pipeline scale inputs?

mixer: Switch between inputs

We can currently add new inputs which sit over top, but it would be nice to trigger which input is on top (potentially pause/remove original input)

Jitter in video when pushing to RTMP

This could just be simply not enough cpu/memory on the transcoder/transmuxer size. Do some more debugging.

tracing

https://crates.io/crates/tracing

mixer: Test RTMP inputs

authentication

We need our API to be authenticated or use mutual tls

input not scaled when added after output created.

Scale input based on mixer dimensions if input doesn't have its own dimensions.

mixer: testsrc resolution

Enforce resolution. It's sized differently based on when other inputs are added.

gstreamer error while copying. Unable to access, server certificate verification failed.

This happens while making an image from Dockerfile.
What to do?

gstreamer probes

We want to be able to collect data like bitrate on input and output and react accordingly

server: Graceful shutdown

We need to store all our JoinHandles from our mixers and wait on all of them (and the http server)

http: Refactor so that we have separate types for Requests vs Response

mixer: Handle pending states when doing state change

It is possible when we try to change the state of a running pipeline. We may end up in the middle of a pending state change that's occuring.

support for inputs/outputs that are audio or video only

input: location & size (this will enable picture-in-picture)

output: Auto freeze when consuming RTMP

mixer: background video is not scaled properly when it's the only video src

refactor inputs into their own bins (intervideosrc/intervideosink)

input: Named states

We could have various mixer::Config's for an input and name them. Then we could swap between them.

PipState
FullScreenState

input: Record to file

mixer: Force watch_bus to shutdown on stop()

mixer: configurable output resolution, colors, volume

docker: Trim down runtime container

Do we need to use nvidia/cuda and the entire Ubuntu image? Can we start from Scratch and copy over any dependencies that are needed?

better error handling

We need to ensure that we capture errors. We can't use things like unwrap() or expect(). We don't want to kill our pipeline or worse all the live mixers for errors.

Add linter to project so it won't compile with unwrap or expect

recording input streams

Depends on #49

Allow inputs to have multiple connections.

We may want to push the same stream to multiple channels/mixers. We may also want to take an input and send it to a file so that we can record it.

video discontinuity when adding an input to an already running pipeline

This is only an issue if an output (this could just be RTMP?) is added before the input is added.

closed captioning

server: Implement custom config

gstreamer errors when running tests

race conditions due to not blocking pads, pending state change, or unlinking when not idle?

Sending bad request breaks app

Try to create a mixer that already exists. We see these errors.

thread 'tokio-runtime-worker' panicked at 'called Option::unwrap() on a None value', src/mixer/mod.rs:251:24

note: run with RUST_BACKTRACE=1 environment variable to display a backtrace
thread 'tokio-runtime-worker' panicked at 'called Result::unwrap() on an Err value: "PoisonError { inner: .. }"', /home/erik/code/switcher/src/http/handlers.rs:204:25

Curl usage. Examples to configure a pipeline please.

I've managed to compile rtmp-switcher and is happily listening on 0.0.0.0:3030.
Is there any examples on how to configure an input=>mixer=>output.
I tried :-
$curl http://0.0.0.0:3030/mixer and got no response?
Am I missing something?

server: Allow command-line arguments

output: Make encoder and encoder presets configurable

Look into IdlePadProbes

We may have issues with changing the pipeline when it is in different states. We want to add an IdlePadProbe to wait till the pad is idle before doing things like unlinking or removing it.

Also look into blocking src pads until ready.

network resiliency

We have not attempted to add latency or jitter to our inputs or outputs to see if the pipeline breaks

https://gstreamer.freedesktop.org/documentation/tutorials/basic/streaming.html?gi-language=c

API requests fail with 500 when mixer does not exist.

These should be 404. Even better with a json error body explaining which component wasn't found