Looking for correct and fast way to overlay an image onto an interlaced video

[renevane]

Hello everybody!

This is my first post in the forum. The past couple of weeks I have been immersing myself in this wonderful piece of software and I still feel I'm only scratching the surface of what is possible!

For now, I am trying to find and implement the correct and fastest way to overlay a transparent .png-image (1920x1080px) onto an interlaced video, de-interlace the video and scale down the result. These are the file specs I'm dealing with:

Input: H.264 mp4 video, 1920x1080px, interlaced, 25 frames per second (so 50 fields per second).
Desired output: H.264 mp4 video, 1280x720px, progressive, 50 frames per second, with the .png image overlayed.

Attached is the Workflow in .json and .png format.

I'm not sure about de-interlacing because there is also a separate de-interlacing processor, which I'm not using because the current output is already progressive. Am I doing something wrong? Also, the whole process is already kinda slow which makes me doubt my current approach.

Can somebody help me with this?


Kind regards,
Rene

[FranceBB]

I haven't checked the workflow yet 'cause I'm using my mobile, but since your image is gonna be overlayed and your final output is gonna be progressive, I would do the following in the following order:

- Watchfolder
- AV Decoder
- Bob Deinterlacer (BWDIF)
- Overlay filter
- H.264 Encoding
- Delivery

Which is pretty much what you're doing already except for the Deinterlacer.
About it, don't worry, it's safe to place it regardless 'cause if your input is progressive, it's not gonna do anything.
This of course if it's flagged as progressive.
If your input is flagged as interlaced and it's actually progressive, a Deinterlacer will Bob but both fields are the same so it will just duplicate frames and you'll end up with 25p duplicated to 50p instead of blended.

[renevane]

Thank you very much for your time Francesco! I now understand that the H.264 processor does not actually do any deinterlacing.
I have now changed the workflow according to your suggestion (with the Deinterlace processor in place) and will work from there.

It is also my understanding that the watermarking proces could be done in FFmpeg through the Command Executer processor to speed up the Workflow, is this correct?

If so, I'll probably be able to figure out what FFmpeg command line parameters to use. But I won't have to go through the trouble if the speed gain would be marginal.

[admin]

Hi renevane, thank you for using FFAStrans and wecome to the forum!

No workflow will ever be faster using the a/v decoder and filters than constructing a special ffmpeg command for the task. So if speed is your top concern, custom ffmpeg is the way to go. But remember that FFAStrans takes care of lot's of other things you would have to worry about when constructing a custom ffmpeg so if it currently works and speed is not TOO terrible I would not change it.

-steinar

[renevane]

Thank you Steinar, much appreciated. I am very happy with FFAStrans and to be fair: currently I'm running it on a HP Proliant DL380G7 with two Intel Xeon X5660 CPU's. In 2010, this was fast. And for Standard Definition video material it does the job. But it does show its age when using HD content.

[momocampo]

Hello Renevane and welcome here!
I have got exactly the same ProLiant Server as you. It's an old system but it works pretty good... for FFAStrans installation files. Yes, about encoding it's a little too old.
Anyway, if you have choice, you should let FFAStrans install files on ProLiant and use hosts(1, 2 PC or more) to process your jobs. In that situation (which is mine), you secure your installation and you can restart easily any host if you need to.
It was just an advice, ffastrans allows several configurations and that's why it is cool.
Feel free to ask questions if you want.
Bye.
B.

[renevane]

Thank you B.! And thanks for your suggestion about the server. Speaking of which: I'm hoping I might be able to invest in a much faster one in the near future. But if not, I will start looking into configuring a FFAStrans farm with maybe two or three PC's to get some more CPU power.

Speed is currently a big issue but mainly because post-FFAStrans tasks (such as uploading finished files to a publishing environment using API calls etc.) have yet to be automated. Somebody (sometimes me ) actually has to wait for a file to be delivered by FFAStrans so it can then be manually uploaded.

It will be less of an issue when publishing is automated. Then everybody can go home on time .

[momocampo]

Don't worry, the more you will use FFAStrans the more you will find possibilities and way to use it. It's difficult to remember how we could do before
About speed, of course it depends on a lot of things (network, workflows, files sizes, etc) and it's difficult to advise you on something but feel free to ask, we gonna try to help as much as we can.
I think almost nobody has the same installation here
Enjoy and see you.
Benjamin.