YouTube Ingest by Hand (Part 5)

I've got the SIDX atom in place at the start of each M4F file, immediately preceding the MOOF atom. No change in the black rectangle.

Hmm.

I noticed in an example MPEG-DASH stream that it has a different set of "compatible_brand" values (avc1, iso5, dash) in the FTYP atom, than what AVAssetWriter outputs (isom, iso5, hlsf). Could be important? We shall see.

No dice after updating the FTYP with the new "compatible brand" values. I really need to find a *working* example of a @YouTube MPEG-DASH ingest stream. Or I need to try making one with FFmpeg and pushing those files to YouTube to see if it works.

I've been unable to find any evidence online that anyone has ever achieved a functioning MPEG-DASH ingest with YouTube. I filed a bug report. I'm throwing in the towel on MPEG-DASH + @YouTube for the time being. https://issuetracker.google.com/issues/277950710

It's still quite possible that the last remaining problem is my failure to correctly mux the separate audio and video fragments, which YouTube claims to require, and which AVAssetWriter refuses to produce.

Change of strategy. The new strategy is HLS. The upside is that YouTube's HLS ingest supports HEVC. The downside is that it uses M2TS instead of BMFF, so I've got another format to learn.

Another HLS upside is that @OBSProject supports YouTube HLS ingest, so there's proof of a working example. It's telling that they don't support YouTube MPEG-DASH ingest. I wonder why....

To be continued.