dont edit in hdv?

[Amos Kim]

I have an hd100 and I hear that one shouldn't edit in hdv that it should be captured in hdv and converted into a loseless format and then edited. What is the best format to convert it to for optimal quality?

[Earl Thurston]

You'll probably have a lot of people recommend Cineform. I'm using Aspect HD in Premiere Pro and feel it's worth every penny. Aspect HD allows me to edit footage from the HD100 in real time on a computer that's below Cineform's minimum specifications (they recommend at least a 2.8GHz Pentium 4; I've got a 2.66GHz). Cineform also has outstanding customer support -- best I've ever seen in any computer-related sector.

[Stuart Campbell]

A decent enough edit station will allow you to edit native m2t hdv on your timeline without issue.

I've an e6700 dual core 2.66 pentium with 2Gb ram and 2 x sata raid 0 drives for video storage. I have no problems editing m2t on the the timeline. However, I choose to convert to avi intermediate using Cineform for capturing purely because it keeps things smooth when using plenty of effects, grading and transitions.

Stuart!

[Brian Luce]

What are some of the other choices besides Cineform? Canopus has HQ that works with Edius. What are the others?

[Paolo Ciccone]

Amos, you can edit in HDV but you should not do anything else. No rendering for export, no effects, no color correction. Get your cuts in HDV, export the sequence to AE or other compositing software and master it at 16 or 32 bits. Output the master using a lossless codec like SheerVideo, Tiff sequence or Uncompressed. That will lead to the best image quality.

[Amos Kim]

Thanks but what if I want to do some color correction or rendering in final cut?

[Dennis Robinson]

Originally Posted by Amos Kim

Thanks but what if I want to do some color correction or rendering in final cut?

Hi Amos,
I am afraid I don't understand the problems editing in HDV. I produce TV commercials for a living and edit everything in HDV.
I am using FCP5 and shooting with the JVC GYHD cam. I find colour correcting and chromakeying is much better in HDV than DV. Once i have the commercial ready I simply copy and paste the project into a new SD timeline.

[Amos Kim]

I think the issue is that hdv is a compressed codec and doing anything to it degrades the image greatly. Using an uncompressed codec improves the image.

[Paolo Ciccone]

Originally Posted by Amos Kim

Thanks but what if I want to do some color correction or rendering in final cut?

You can but you can degrade the image quality. FCP is not the best tool for that job. Same thing for all NLEs, that's why at the high end people edit in FCP or Avid and then Color Correct on something like DaVinci. I did a lot of tests, tried all kind of solutions, at the ned I get the best quality and flexibility, a pretty important issue for me, in simply importing the sequence in After Effects and completing the job from there.
If you want to use FCP I would suggest at the very least to switch your sequence to high precision YUV, use Magic Bullet Colorista instead of the CC 3 way and set your sequence to use Uncompressed as compressor or SheerVideo.
In addition you might want to try to convert you footage to SheerVideo or other higher precision, lossless codecs, before editing in FCP. It all depends on what is the weight of color fidelity versus time or other constraints. Only you can judge that :)

[Amos Kim]

Thanks for the advice paolo... but can sheervideo be used export between tape/vtr and fcp? Or does recording to HDV tape already compress irreversibly? How does it compare, in your opinion, to AIC or other lossless codecs?

[Paolo Ciccone]

Amos, once you record to tape you use HDV compression with 4:2:0 color sampling. Nothing to do about it, it's the nature of the beast. You can record to 4:2:2 by using component out, which gives you marginally better color, more sampling, but you still get heavily compressed, 8-bit per channel, color. This is what we accept when using an HDV camera. Short of using RED there's is pretty much no alternative in this price range and it's not a problem if you avoid any subsequent loss of data.

The main problem of HDV is that it's a temporal compression. GOPs are created by storing frames as deltas from a keyframe. When editing you end up cutting inside a GOP. If you output the sequence back in HDV or do any work that requires rendering in HDV you automatically recompress you image because the software needs to create new GOPs. If you do CC in 8 bits you often end up with banding and other artifacts caused by clipping the signal. This is easier to do than it seems. CC should be done always at at least 16 bits. In AfterEffects this is rather easy, in FCP not so much. That is probably the reason why you need to render from Color.

SheerVideo is a lossless codec, that means that when you uncompress the frames they return to exactly the original state. They are bit-per-bit accurate. AIC and ProRes and Cineform are not lossless codecs. They claim that they are "visually lossless" but the distinction is necessary. They do throw away information. The developers of those codecs claim that the loss is not visual, in other words you can't see it. That might be but as soon as you start recompressing the clips a few times, for ProRes it's been estimated in around 6-7 times, you are able to spot the difference by naked eye. This means that the data loss is there. Tim D. and others posted some results time ago when AIC came out and there was general agreement that AIC was decent but you could spot the difference by doing an accurate test.
Even when you cannot spot the difference chances are that your CC software will behave differently and will cause artifacts at some point. This could be entirely acceptable, for me it usually isn't.

but can sheervideo be used export between tape/vtr and fcp

It can. You can download the free trial from http://www.bitjazz.com . It's fully functional with no watermark for 20 days. The software install itself as a QT plugin so you can convert your m2t files into Sheer by using MPEG Streamclip and then loading the resulting .mov files in FCP. Or you can import from FCP and use Sheer as the sequence's compressor. The reason why I recommend it so much is that it's about 1/2 the size of Uncompressed but it runs generally faster than Uncompressed. This means that you worry less about the storage requirements of HD footage and you get actually better performance from your machine.

[Amos Kim]

Wow, that makes total sense. Thanks Paolo!

[Brian Luce]

Originally Posted by Paolo Ciccone

It can. You can download the free trial from http://www.bitjazz.com . It's fully functional with no watermark for 20 days. The software install itself as a QT plugin so you can convert your m2t files into Sheer by using MPEG Streamclip and then loading the resulting .mov files in FCP. Or you can import from FCP and use Sheer as the sequence's compressor. The reason why I recommend it so much is that it's about 1/2 the size of Uncompressed but it runs generally faster than Uncompressed. This means that you worry less about the storage requirements of HD footage and you get actually better performance from your machine.

Interesting. I haven't heard of this codec yet--has anyone done any comparisons with Cineform? pros and cons?

[Paolo Ciccone]

Comparison of quality is not an issue because Sheer is completely lossless, what you get from you camera is what you keep. It would be unfair to compare against any lossy compression because pixel-by-pixel there would be differences.
You can re-render SheerVideo clips as much as you want, you are not going to have any generation loss. Size-wise it's larger than Cineform, there is no competition for that but I want say that just added another 500GB Western Digital hard disk to my storage for $160.00. Storage is really not an issue. Sheer compresses at less than half of Uncompressed but runs faster.

[Steve Mullen]

Originally Posted by Paolo Ciccone

Amos, once you record to tape you use HDV compression with 4:2:0 color sampling. Nothing to do about it, it's the nature of the beast.

The main problem of HDV is that it's a temporal compression. GOPs are created by storing frames as deltas from a keyframe. When editing you end up cutting inside a GOP. If you output the sequence back in HDV or do any work that requires rendering in HDV you automatically recompress you image because the software needs to create new GOPs. If you do CC in 8 bits you often end up with banding and other artifacts caused by clipping the signal. This is easier to do than it seems. CC should be done always at at least 16 bits. In AfterEffects this is rather easy, in FCP not so much. That is probably the reason why you need to render from Color.

Color is a separate program that originally -- pre Apple -- HAD to render its output just like AE. If you wish, you can render to HDV. But don't. Choose ProRes 422.


Yes -- HDV is 4:2:0, and nothing you do later changes that. The decimated chroma information is lost and no CODEC you use post camera fixes this.


During editing in FCP, every MPEG-2 frame -- it makes no difference where it is in the GOP -- is "constructed" from the appropriate "other" frames. It is a frame built from motion vectors -- not delta frames as is often wrongly described. It is a frame with full information. (Step through MPEG-2, the quality is equal for every frame.)

Next, FCP converts each frame to 4:4:4 uncompressed video. At this point it is no different than a frame from ANY other codec. Whatever information was recorded by the HDV camera is present. No more -- no less.

ONLY if you MANUALLY render, is another codec used. With FCP 6.0.2 this codec is ProRes 422. (Many folks are unaware of the advance introduced by FCP 6!) ProRes 422 is full capable of carrying the information from HDV. There is no value to using a different codec anywhere in editing HDV.

At Export -- almost never back to HDV since one can't really "deliver" HDV to anyone -- the export codec encodes 4:4:4 uncompressed video. This uncompressed video comes from: decoded HDV, decompressed ProRes 422, or uncompressed graphics.

Thus, the only codec induced loss is HDV in the camera; in ProRes; and in your choice of export codec. You'll almost always be exporting uncompressed via HD-SDI to HDCAM or DVCPRO HD. So the export "render" is done by your VTR. (Note, FCP renders to HDV only if you record to HDV tape.)

Compression in ProRes is done only ONCE so there is no issue of generational loss. Renders are NEVER reused.

ProRes 422 HQ at 10-bits is able to perform as well as anyone who starts with HDV could possibly need. :)

Moreover, both HDV and ProRes support real-time editing! This is crucial! And, not true of other codecs such as CineForm.

Introducing another codec is needless as it gains no quality increase. Editing anyway but native HDV is a total waste of disk space and bandwidth. And, money.