I'm an experienced developer but a complete newbie to Premiere Pro so forgive my ignorance on PP specifics. I've been asked to look at the options for integrating Premiere Pro with an in-house video repository which exists outside of the local file system. We have a REST API for querying metadata and video references, and in practice the physical video files reside in AWS S3 buckets. Many of the video files are into the hundred of MBs, sometimes GBs, so it would be prohibitively slow to download a file from S3 before opening a local copy (or to upload it after saving). What I'm hoping is that it might be possible to write a Premiere Pro extension which would allow us to pass video data to and from S3 as required rather than dealing only with the local file system. For example if Premiere Pro is playing a 1GB video and the user moves the play position to the halfway mark than I would hope that the application might be able to request the extension to read data from file offset = 500MB, allowing the extension to set the S3 stream offset to the same point in the remote file.
Is such a thing possible using either the plug-in SDK or with a CEP panel?
From what I've read in the two SDKs so far:
- As a plug-in it looks like this would be an importer type, and I can see that it should be possible to have an importer which supports the video formats we're using (in practice likely to be mpeg4/h264). However it looks like an importer really works at the frame level, which we don't want to get into. We'd really want the plug-in to act as a sort of proxy layer for the file system, with Premiere Pro calling the plug-in to read or write data, but the data being in an existing known format which the plug-in doesn't want to be decoding or encoding. However that doesn't seem to be an option for an importer.
- I can see that a CEP panel could provide the means to query our external repository and then pull a video into Premiere Pro. This sample code does something a bit similar, querying Flickr then loading a selected image into Photoshop. However that code pulls the whole image down from Flickr to the local file system then pushes an openDocument request into Photoshop. That's not going to be an option for us with such large video files in S3 and when the application may be used in satellite locations with limited bandwidth.
Have I missed something in either API which would allow this sort of "data redirection"? Or does anyone have an alternative approach to the problem?
Thanks for any help.