Resolume Forum

What is the technical reason for why DXV-alpha creates videos which are nearly half the file size compared to that of HAP-alpha?

When comparing the rendered results between DXV-alpha and HAP-alpha, the edges of the alpha are slightly different and HAP-alpha better matches the original footage. But it's not enough to notice while VJing and I can see why it's likely a purposeful decision by the programmers behind the DXV codec.

I suspect that the DXV-alpha codec is throwing out some information in the alpha channel at render time, but I cannot find any documentation (or discussion) on the subject, and the DXV codebase is private.

Could a Resolume developer shine some light on this?

File Size vs quality is always a compromise.
Try DXV HQ, it should be more accurate.

I'm currently doing research for an article since I frequently see VJs that are curious to understand the differences between the DXV and HAP codecs. But because the DXV codec is closed source there is not much known about the implementation details. So I've rounded up any info I can find into the article linked below. Any knowledge you can share would be greatly appreciated.
https://www.jasonfletcher.info/vjloops/hap-vs-dxv.html

TECHNICAL QUESTIONS

• Why is there such a big difference in the file sizes between DXV-Alpha (1.20 GB) and HAP-Alpha (2.63 GB)?

• I would think that the DXV-Alpha file (1.20 GB) would be larger compared to the DXV file (1.37 GB), due to the addition of an alpha channel. But that's not the case. Why?

• What is happening at a technical level to the alpha channel when using the DXV-Alpha or DXV-HQ-Alpha codecs? This would help to explain the visible difference in the edges of the alpha when comparing to the source footage.

TEST RENDER: INPUT
- 1 minute of footage at 3840x2160 60fps with alpha

TEST RENDER: OUTPUTS
- DXV: 1.37 GB
- HAP: 1.61 GB
- DXV-Alpha: 1.20 GB
- HAP-Alpha: 2.63 GB
- DXV-HQ: 1.36 GB
- HAP-Q: 3.20 GB
- DXV-HQ-Alpha: 1.55 GB
- HAP-Q-Alpha: 4.06 GB

We are not prepared to share such details at this time. Not only because DXV 2 & 3 are closed source codecs, but also because we dont know. When developing DXV 2 & 3 we made choices we think are best for VJs specifically. We dont look at or 'borrow' technologies from other codecs, so we dont know what the technical differences are.

HAP has reverse engineered our DXV 2 codec and implemented a near duplicate. They aren't done reverse engineering DXV 3 or the alpha mode it seems. There's no reason for us to help them do that.