New H264/5 Export settings?

DaVinci Resolve seems to have recently added some new export settings which look similar to OBS's recording settings.

The bitrate section is the same, but the new settings seem to be:

Encoder: now has an auto option (Auto by default)

Preset: Different speeds listed (Medium selected by default)

Tuning: High Quality or Low Latency (High quality by default

Two Pass: Disabled, Quarter Resolution, Full Resolution (Disabled by default)

Should I change any of these from the H265 Master defaults?

Yes.

SDR Deliver.png (63.97 KiB) Viewed 6835 times

HDR Deliver.png (62.88 KiB) Viewed 6835 times

Yes.

SDR Deliver.png

HDR Deliver.png

Interesting. Why H.264 for SDR delivery? Shouldn't I use H265 for both since it produces higher quality at smaller file sizes?

I do it mostly for standards compatibility. H.264 is pretty ubiquitous for HD/SDR, and H.265 seeming to the the norm for UHD/HDR.

No other reason.

I do it mostly for standards compatibility. H.264 is pretty ubiquitous for HD/SDR, and H.265 seeming to the the norm for UHD/HDR.

No other reason.

That makes sense. For YouTube, then, the smaller file size should help with upload times and the ability to render to h264 will be good if it’s required for something paid.

I only discover this change now after upgrading to a newer version.

Does this affect the presets that I already made?

I had a preset that actually exported high quality h264 with a bitrate setting of 350.000 but that setting is nowhere to be found now.
Does this mean I have to reinvent the wheel again? (in other words: spend another day trying to get a high quality h264 export without weird keyframe artifacts etc.) ?

EDIT: Yes it does mean exactly that. I select my high quality preset and Resolve poops out a video of less than a GB and lots of compression artifacts. The intentions are good of course, and resolve is only getting better and better, but you can't just dismiss carefully made presets. I'm sure many users don't even realize that this is happening until it's too late and they've got and angry client.

Is there perhaps a virtually lossless h264 4k default preset now that wasn't there before? I don't want to spend another eternity trying to find the right settings which have now become so complex you need a masters degree to understand..

Export in a mezzanine codec and then encode with HandBrake.

Export in a mezzanine codec and then encode with HandBrake.

I've never found that necessary and it only adds an extra step to the workflow.

For my purposes, rendering straight to H265 using NVIDIA's engine is more than good enough.

I found Apple’s encoding good enough too.
But if you want more control, HandBrake is a good alternative.

I don't want to spend another eternity trying to find the right settings

Use the ones I posted above.

(Changing only Resolution, Frame rate and Key Frames appropriately.)

What key frames do I need for 25fps? Still 12? Or automatic?

LOOKAHEAD option now includes levels 0 to 3, which is the best level for high quality 4K 24FPS HDR?

I found Level 3 to be the sweet spot for me.

I found Level 3 to be the sweet spot for me.

I'll have a play, just left it at 0 so far

I've noticed they've just added split mode too to the export options. Just Split/Auto options, what on Earth should that be set to?

I haven't found that answer yet.

LOOKAHEAD option now includes levels 0 to 3, which is the best level for high quality 4K 24FPS HDR?

See https://docs.nvidia.com/video-technolog ... head-level

Edit: I just tested and my 3060 TI only supports Lookahead level 1.

I've noticed they've just added split mode too to the export options. Just Split/Auto options, what on Earth should that be set to?

Split encoding makes Resolve encode to 2 Nvidia encoders at once which can increase speed of encode. I recall reading that Nvidia themselves say the quality is lower by using split encoding. Would be interesting see a quality analysis between split and non split encoding.

4070ti and above 40 series have the dual encoders

NVIDIA Ada Lovelace Archite.png (113.29 KiB) Viewed 2126 times

Yes.

SDR Deliver.png

HDR Deliver.png

One typically does not want constant QP; CRF is great (but hardware implementations typically cannot handle this). In short, I would not use constant QP.

One typically does not want constant QP

Well, where quality is more of a concern than size...

So using Simon's settings for high res 4K, which I do, Split will be bypassed even if selected?

These are my typical render settings for YouTube. Original footage is HD upscaled with Super Scale 2X to UHD in a UHD timeline. This 2min 37sec render came out at 1.53GB, though I never worry about final file sizes.
In case anybody wonders, the timeline resolution is my preferred 2.25:1.

Split will be bypassed even if selected?

Looks like.

Export in a mezzanine codec and then encode with HandBrake.

No. Not Handbrake.
Use Shutter Encoder for a good result.

I like Shutter Encoder just as much.
But both are just frontends for ffmpeg, so identical settings will deliver the same quality.

One typically does not want constant QP

Well, where quality is more of a concern than size...

QP does not consider how complex a scene is when compressing video. As a result, the quality of the footage can vary a lot. Complex scenes may look worse because they get the same level of compression as simpler ones, even though they need more data to look good. This inconsistency leads to some sections of the video showing noticeable flaws during detailed or fast-moving scenes, while static or simple parts look better than necessary. This rigid approach to compression makes QP less suitable for videos with different scene types, like movies, sports, or gaming, where it's important to keep the quality consistent throughout.

QP does not consider how complex a scene is when compressing video.

That's exactly what it does. More complex, more bits. No limits to how many bits as with CBR and VBR.