Ok, that makes sense. I tend to do VFX on ungraded material because it gives the director/DOP/colourist the latitude to change their minds. It can also help when there are multiple colourspace deliverables to be made, theatrical DCP in XYZ, Home Ent in HDR, Home Ent in SDR, etc; If it is a music video though, I doubt you will need to account for theatrical. Out of interest, do you know how they are deriving their SDR? Are they doing a Dolby Vision workflow, are they grading in ACES and simply doing an output transform, are they planning on doing a SDR grade from the HDR source, or are they doing a new grade from the original source (yuck, not sure why anyone would do that but I know it happens a bit by ‘colourists’ who don’t really know what they are doing).
Unfortunately, the way the colour house is doing the grade will dictate the best workflow. I can breakdown the 3 most likely though.
If ungraded, either work in the native log colourspace or work in ACES using exrs. Do NOT ever get Log provided footage in exrs as this is destructive. I’m all about working in ACES, transporting it in AP0 ACES2065-1, working in ACEScg, then delivering back as ACES AP0. The beauty of ACES is you can view it however you want at your end, just view through the output transform that suits your monitor. There are plenty of good resources on Logik for working in ACES so no need to elaborate. Working in LOG is equally valid, but try to work uncompressed or with minimal compression in a bare minimum of a 12bit codec. Definitely DO NOT work using any 10bit codecs as this will destroy the image when grading HDR then deriving SDR from that.
If working on graded images, hopefully they are grading in ACES. If so, get the graded pictures sent in ACES AP0 in exrs, do your work once again in ACEScg, viewing through whatever output transform suits your monitor, then send back to them in ACES AP0. The beauty of ACES workflows is you can pretty much work in any manner on whatever system and tailor your output transform to whatever display you have got at hand. The HDR P3D65PQ, Theatrical P3CDI or P3D65 at gamma 2.6 and SDR all translate well from an ACES grade. If you are working on HDR graded pictures this would be my recommended workflow.
If they want to send you the HDR grade in P3D65PQ that is also fine. Once again, ensure that you are working in a minimum of a 12bit codec. As long as you tag it correctly in Flame, then you should be able to display it correctly. If you only have SDR monitoring, a quick trick is to add a Dolby HDR node and do an analysis and that will give you a good represenation in SDR. If you have an XDR display, it should do a decent job of displaying the P3D65PQ accurately anyway, just make sure you are viewing it correctly in Flame. Is your GUI the XDR though or are you using it as a broadcast monitor?
If you need to create approval files from the HDR, you will need to know how they will be looked at. If they have a decent HDR display (such as an iPad Pro) then you can just send them the graded HDR. If they only have a SDR then I would suggest using the Dolby HDR tools within Flame to analyse the shot and create the SDR deliverable using that. ACES is also an option for this using Colour Management within Flame.
Feel free to contact me directly once you have chatted to the colour house. The way they want to grade the music video will dictate the way you work on the shots. Personally, I would always prefer to do the compositing on the ungraded but working on graded HDR images is equally valid.