S3 is not a block file system, it is an object file system. The difference here is that, normally, you cant mount a S3 bucket like a standard unix FS and work on file with fopen(), fwrite() ect... Some trick exists to work on S3 like any other FS but I would suggest an other option :
You have to transcode the video on a localy mounted FS (like an AWS EFS, or a local file system), then "push" (or upload) the whole transcoded video onto the S3 bucket. Of course, you can improve this process in may ways (remove temp file, do parallel works, use Lambda service, or task in containers...). You should avoid to do many upload/download from or to S3 (because it is time and cost consuming). Use a local storage as much as possible, then push the resulting data when they are ready on S3.
Also AWS have a service to do video transcodification : https://aws.amazon.com/en/elastictranscoder/