Using cameras

[leiradel]

My old implementation was already getting frames from the camera in a separate thread. Let's see if things are better with the video module.

I don't need to modify LÖVE, just to extend VideoStream and return an correctly-filled frame in getFrontBuffer(). This may require some massaging in the framebuffer provided by the camera library...

[bartbes]

PS: This module is not encapsulted at all, and right now is only designed to play Theora Ogg videos... We can ask the developers to make this better but don't expect much though. I would love to see a C API to implement something in the C side that can behave as VideoStreams or something like that.

I'm not sure what you define as "encapsulated", but it's not tied to theora at all. To implement an alternate decoder, you simply need to implement a VideoStream (and its base class Stream). The trickiest part is probably seeking. During development I had it stream from Video4Linux, after I wrote an ffmpeg-based decoder, LVEP, that can be built and required like any lua module can.

love.graphics doesn't need any modifications at all, provided your video output is in Y'CbCr format (as VideoStream specifies).

[Positive07]

Oh! Sorry I thought that there was no way to define the same structure and change the functions, as I said I have no experience with C++ and it's classes. I'm glad this is possible.

The video output format may be a problem, the seeking part could totally be ignored with a dummy function or something, since seeking camera input is not something you would expect, unless you are recording or something, but I wouldn't go that far for a first implementation, and if I did I would do it in parallel and directly to an ogg file with a separate stream so that you can save that file if needed...

What I don't have any idea how to implement is the sound part and syncing if you can provide more insight on this I would love to learn about these topics

Thanks for your answer bartbes, I'm glad you came to correct me and I learned new stuff

[bartbes]

Sound is basically the same, but then using Decoder. Same deal really, except sound is not double buffered (it's an older interface and it hasn't been updated to match). As for syncing them up, a VideoStream has an associated FrameSync object, which it is supposed to sync to. This FrameSync can either just be time-based (DeltaSync), or associated with a Source (SourceSync), in which case it uses that Source's playback position as the current position.

Unfortunately, for a camera, I'm not sure how you would sync it up to anything, and if there's even information available about when a frame was recorded, and when the relevant audio was. Again, since FrameSync is (designed to be) a base class, theoretically you could define another synchronisation method.

[alejandroalzate]

there is a way to get this working with a udp stream from a wifi camera?