Who went from the DVX to a DSLR?

HVX vs 7D

HVX vs 7D

First of all, there is no "vs". Just finished a production where I used the HVX200 for all the interviews (backlit white silk background) and 7D for all the (720 60P) b-roll slow-mo. Both have cameras have distinct characteristics and edit perfectly together. Next production I may use the 7D for interviews (shallow depth of field) and the HVX200 for b-roll. It'll be a long time before I get rid of either one - they go well together as long as you realize each one's capabilities.
 
Michael Carter said:
Nobody should be editing H264, it's a delivery codec. Come on, this isn't even an issue.


My new workflow: copy the card, open shots in streamclip, trim and send to batch. I find it dramatically faster than the DVX method.
Click to expand...

I hear this alot. (and don't really understand it) I shoot, copy the card to the box, put the clips on the timeline, sync audio, cut/trim, whatever effects, export the end product to whatever format. Why can't you edit in H264?
Granted, I've only had the T2i for a month or so, but I've never had a problem.
 
Because H.264 is going to lag your computer a bit. Even if you have a fast machine and don't notice any slowdown, it's always going to be faster to edit transcoded footage. Additionally H.264 won't handle effects like color correction, image stabilization, or general compositing nearly as well. Using it will also increase all your render times. You might not notice at each step, but it quickly adds up to much more time than transcoding takes.
 
diego1235467 said:
I simply would not trade in my DVX because it's a VIDEO CAMERA, not a still camera with video capabilities.
And honestly, as nice as footage from DSLRs look, I always find something off about it. I think DVX by far gives a much more cinematic look.

In my mind, the DSLR is for shallow DOF, low light, music videos, commercials, etc. A DVX is for deep DOF, handheld, short films, etc.
Click to expand...

i agree with this which is y we kept the dvx as well. were currently doing filming on a documentary using it
 
+1

DVX and a new T2i in the kit right now. Bit of a learning curve going on but I'm loving the form factor and the ability to leave the RR cine adapter home.
 
Ben_B said:
Because H.264 is going to lag your computer a bit. Even if you have a fast machine and don't notice any slowdown, it's always going to be faster to edit transcoded footage. Additionally H.264 won't handle effects like color correction, image stabilization, or general compositing nearly as well. Using it will also increase all your render times. You might not notice at each step, but it quickly adds up to much more time than transcoding takes.
Click to expand...


I respectfully disagree.

Canon DSLR's store video in H.264 codec inside QuickTime wrapper. This being a compressed lossy format, any further transcoding to anything else but loseless will degrade picture. Thus, one should try to edit H264 directly (yes, it takes CPU, thank you Intel for i7 CPU's!) ad output directly to whatever delivery format is used.

My rig plays two native H.264 1080p in real time, editing is a breeze. Once project is cut, it gets printed to whatever is to be used.

This way, amount of transcoding is kept to minimum.

If we break it down to details it looks something like this:

1. H264 -> RGB
2. RGB -> transitions, cuts, grading, whatever
3. Out to output (BluRay, RAW, H264, MPEG2).

If you transcode to anything else but loseless you will add one further lossy transcode on the top of H.264 one.

Basically, native H264 editing means that every frame will be natively converted to raw, done whatever is to be done and then compressed again (losing some quality). Of course, decompressing frames from highly convoluted codec as H.264 in real time costs a lot of CPU juice. But I firmly believe that pipeline should be kept as short as possible and all post-processing should be done on raw data.

There will always be less information in transcoded lossy file than original H264 one. Transcode is a photocopy of original and original already throws away 90% of information. Thus, you cannot expect transcoded file to suffer grading better.
 
Grunf said:
I respectfully disagree.

Canon DSLR's store video in H.264 codec inside QuickTime wrapper. This being a compressed lossy format, any further transcoding to anything else but loseless will degrade picture. Thus, one should try to edit H264 directly (yes, it takes CPU, thank you Intel for i7 CPU's!) ad output directly to whatever delivery format is used.

My rig plays two native H.264 1080p in real time, editing is a breeze. Once project is cut, it gets printed to whatever is to be used.

This way, amount of transcoding is kept to minimum.

If we break it down to details it looks something like this:

1. H264 -> RGB
2. RGB -> transitions, cuts, grading, whatever
3. Out to output (BluRay, RAW, H264, MPEG2).

If you transcode to anything else but loseless you will add one further lossy transcode on the top of H.264 one.

Basically, native H264 editing means that every frame will be natively converted to raw, done whatever is to be done and then compressed again (losing some quality). Of course, decompressing frames from highly convoluted codec as H.264 in real time costs a lot of CPU juice. But I firmly believe that pipeline should be kept as short as possible and all post-processing should be done on raw data.

There will always be less information in transcoded lossy file than original H264 one. Transcode is a photocopy of original and original already throws away 90% of information. Thus, you cannot expect transcoded file to suffer grading better.
Click to expand...

What codecs have you been (not) using? Transcoding to a professional codec like ProRes will not degrade the picture.

Although like any generation you are throwing data out, ProRes is designed to be visually lossless. You can go generation after generation with ProRes without losing any quality, unlike compressed codecs like HDV and H.264.

H.264 is designed as a delivery codec and now is being used as an acquisition codec. It was never intended as an editing codec. ProRes is designed to be an editing codec, and works well as that. Color grading, keying, compositing, etc, will all work better with the ProRes footage, and editing will be faster.

Transcoding to ProRes is not like photocopying the original it's more like translating binary code to normal English. You can't read binary as well as you can English (I hope) so it will be slower. Yes there will be data lost (fewer digits in a long chain, requiring more data to be squeezed through every second) but no information is lost (the picture quality is the same, the words are the same.)

http://www.larryjordan.biz/articles/lj_understanding_ProRes_422.html

Read this and take a look at the chart comparing HDV to ProRes. All of these comparisons hold true for H.264 except for rectangular pixels. H.264 is a long GOP codec, it is better to edit in an I-frame codec, H.264 is 8 bit, 10 bit is better, H.264 is 4:2:0 sample, 4:2:2 is better.
 
Last edited:
Grunf said:
Thus, you cannot expect transcoded file to suffer grading better.
Click to expand...

It's long since been firmly established that transcoding results in better footage if you are color grading. Or even if you're not. Check around these forums if you don't believe me, I won't do it for you.
 
diego1235467 said:
From what I've seen, a T2i or other DSLRs are far better than a DVX in low light. In fact, I think the DVX is pretty terrible in low light.

I haven't changed my DVX and don't plan to. It's a far better video camera than a DSLR. That said, I still plan on getting a T2i, but never to replace my DVX.
Click to expand...

You might want to rethink that. I thought the same way you did. Then I bought my T2i. Even as the cheaper, smaller brother to the 7D, it murders anything the DVX was able to do.
 
Jester2138 said:
It's long since been firmly established that transcoding results in better footage if you are color grading. Or even if you're not. Check around these forums if you don't believe me, I won't do it for you.
Click to expand...

So, you're saying that if I transcode, my footage will somehow become better?
I'd like to meet whoever firmly established that.

While I'm sure that transcoding to something might make editing a little faster for someone with a slower machine, and it might even make color grading stick better after further transcodes. (although I've never had a problem)

Trancoding will never make your footage better.
 
Transcoding to anything else than RAW lossless codec will raise the entropy, period. Editing in H264 directly lets computer transcode frames "on the fly" to RAW frame buffer. As this is somewhat heavy work (due to H264) it won't work on slow computers. But as long as computer is up to speed, it's always better to edit directly than transcode to yet another lossy codec, edit, then code it once again into output format.

If source was 4:2:0, no intermediate format can magically deliver chroma information that wasn't there to begin with. As frame buffers work on 4:4:4 data, I see no point in transcoding to lossy except to speed up editing.

IMHO, only intermediate format worth transcoding to is lossless (either raw or LZW-compressed). Stream is around 150-100MB/sec for 1080p25 which is attainable with two-three harddrives in RAID-0 configuration.
 
Again, ProRes 422 is a visually lossless codec. This has been proven scientifically by Apple during its development, and it is used as such by professionals.
 
The point isn't magically adding new picture information to a crappy codec, the point is giving yourself more headroom for fades, color correction, composting, whatever... For example: Sure, the 8bit codec only has so many steps between red and blue. But convert it to a 10bit codec and start color correcting and suddenly you have a whole lot more steps you can use...

Besides, whether or not it improves picture quality, you certainly don't LOSE any, and you save loads of time dealing with the annoying H.264.
 
Jester2138 said:
Those two are contradictory.
Click to expand...

They do appear to be when you take them out of context a little bit. (but I think you know what I meant)

But to clarify....
Transcodeing will NEVER make your footage better.

I even just bought a new editing suite that has presets for editing in the dslr format.
 
Transcoding will not make your footage "better";

But grading, transitions, effects - most of them add information to your footage. Going with a codec that's designed to deal with those additions will give you a better final product. It may be so subtle as to be unnoticeable - but in the case of grading, from mild to extreme - it's often the difference between a successful grading job or a failure.

A good analogy is retouching 8-bit images; for large final use or for serious color correction, the first thing you do is convert to a higher bit depth. Your final delivery will be 8-bit, but the ability to do smooth changes between colors, and to really finesse the image? It's the difference between (higher-paying & career building) professional work and "hey, I have Photoshop". (And in the case of Photoshop, higher bit rates slow your workflow - at least ProRes speeds things up...)

Are the people that are loudly supporting H264 editing making a living with this stuff?
 
Thanks Michael you explained it much better.

Yes, transcoding doesn't magically improve your original footage with a simple conversion.

But it will ultimately improve the final films image quality once you start editing. You're saying that yourself, anyway, videohog, you just don't realize it.
 
In any case, while some of Grunf's assertions may be correct his conclusion is incorrect unless his goal is simply to save the time it takes to transcode, regardless of the problems.
 
IMO, totally worth the time to transcode. Working in Prores 422 is wonderful and my computer can't really handle H264 anyways.
 
You must log in or register to reply here.
Back
Top