Thats fine ! I just have to make sure what I want is not in these gops any idea how big they are time wise ? ie what is the smallest time roughly that I should pad clips with to prevent recompression ?

ALSO what software on the PC will do this ? ie only recompress what it must when I cut up clips ?

Chris Taylor
http://www.nerys.com/

Thanks, that is good news, that it is not trying to do cheap compression tricks.

I wish Ambarella would hurry up, no information for the March release they wanted. Probably some production delay.

New In camera Edting idea
 

I wish they would add better in editing function in camera. As you know, many computers lack the power to edit H264, Mpeg4, in realtime. What use is it going to be for consumers, if it takes so long to edit on their PC's. They could write an application that allows a pc to control in camera editing, and display live views USBed from the camera, in the editing window. This could work on any PC that can just display H264/Mpeg4, if there was enough power in camera, to change and transmit the view in an intermediary format, then most computers could be used to control the program. This would be a great boon to average users buying it, because it would discourage others from telling them that it is too slow to edit.

Wayne,

I think you've hit on a GREAT idea!

I see most things done in the computer - files stored on the HD, etc.

But using the hardware encoders and decoders in the camera would REALLY help things along!

This would take cooperation between camera makers and editing software makers. Could the codecs in the camera be accessed by the software just like any other software codec? It boggles the mind!

Hardware Encoder
 

The issue with coporation between camera and software manufactures might be dificult. ;-)

Maybe an alternative might be somebody manufacturing a firewire device which is a straight stand alone encoder setup for being controlled by software from the computer.? Plug it in like a dongle and start editing.


Bo

www.bophoto.com/panos

Chris,

I too wonder about GOP structure in the HD1. I have software that allows me to see GOP structure for MPEG2, but nothing for MPEG4. Conversion program from Nero defaults to GOP range from 5 to 300 frames. Apparently the software can dynamically change the structure for greatest efficiency.

I did notice that a test edit I made in the camera did not seem to be too precise. I suppose that the ability to cut a clip in-camera might be most valuable to remove most of a long clip, on vacation for example, to save space. I know I have shot long scenes waiting for a whale to emerge from the water, or a big chunk of ice to drop from a glacier. It's a shame they did not include the feature whereby you start shooting 5 seconds BEFORE you press the start button. It's pretty easy to do with memory card.

I envision a continuing progression of codecs as technology improves. The beauty of using the in-camera codecs is that it's exactly what's needed for YOUR camera.

Actually how much memory would the camera need to have internally to allow that (precatch) and how much processing power since it would have to be a "running" softspace

lets say you wants 5sec that would entail at least 8+mb of very high speed space so it could dynamically write to and erase from this space and then also enough space to act as a BUFFER since once you record it has now 2 jobs.

Record from this second onward retain the last fix seconds DUMP those 5 seconds to the SD card while also dumping the real time recording once you pressed the button and also MERGE the two files seamlessly.

I imagine that is why they did not do it. ie it would probably take a ton of horsepower to do this in hardware on the camera itself.

and if you gonna buffer might as well buffer more I would love 10-15 seconds.

Chris Taylor
http://www.nerys.com/

I wonder about the possibility of using GPUs for editing mpeg stream. GPUs these days from ATI or nVidia should have enough horsepower to handle mpeg in realtime. It's just that there's no software that takes advantage of it. Now mpeg2 editing is becoming more popular with the rise of HDV, something must be in the works. It's gotta be.. I hope :)

For now, on the mac, I could edit HD1 footages I downloaded from here in iMovie HD with ease. Granted it's not in native mpeg4 format but converted to Apple Intermediate Codec, but the fact that I could edit HD size clips on my 3 yr old laptop is a delight. Footages looks 'better' in iMovie than in Quicktime player because 1.25 G4 processor was having tough time decoding mpeg4 in realtime yeilding lots of blocky pixellation.

I hope cameras like HD1 or sony's HC1 and HC3 could bring HD content creation to the mass much faster than expected. Sanyo has potential be the star of that revolution. I hope they hurry and develop the next gen product sooner than later.

oops same post went up twice. editing the second one...

Chris,

I think it could be done more simply than that. I have a Comcast DVR box that constantly buffers TWO tuners to disk. In SD it buffers close to an hour, but in HD it's only about 15 min. on each tuner. If sometime after a program starts, I decide to record it, the buffered stuff is instantly converted to recorded program. It might lead to some fragmentation, but it apparently works.

not it wouldn't really help, because of bandwidth constraint over USB for uncompressed footage in high resolution... You need 30Mb/s sustained for 1280x720 so usb 2.0 can handle that but not more.

The problem is that the actual bus used for graphics transfert aren't designed to _retrieve_ data from GPU. AGP is almost unidirectionnal and PCI-Express, while it should have ameliorated things, didn't do that much.
(not mentionning that gpu aren't that fast for some operations you find in typical mpeg application).

Spending time optimizing things for the cpu is a better choice, imho. There is some freaky fast H.264 decoder (CoreAVC) out there that can handle 1080p footage with my Athlon 1600mhz...

Btw, i'll have a little surprise for HD1 owners tomorrow... it's called MP42AVIEdit ;)

Hey david some issues

first most DVR's use mpeg2 ! (the newest DISH box I believe will use mpeg4)

Second these things are basically full fledged computers. not the tiny little thing your holding in your hands :-)

BIG difference.

The problem is not can it be done but can something this small do it and still be affordable. Remember it has to work in FASTER than real time since it has to manage the buffer AND work the new incoming stream flawlessly. if there is a blip in your recording of lost its no big deal if its a blip in your home video and that blip COVERS exactly what the buffer was supposed to capture .....

thats gonna take a rather larger buffer and a rather effecient and powerful engine to "process" all that data. over 1mb a second is nothing to scoff at.

I guess one way to measure gop is to see how small a clip it will let us cut up ?? Hang on I will try it.

Ok I did a roughly 2 second clip and tried to cut it in half. it seems to have worked fine. so it must be smaller than 30 frames right ?

I have a bunch of clips I am going to upload when I get home tonight I will upload this test clip as well

Chris Taylor
http://www.nerys.com/

What would make me really happy as an HD1 owner ?

2 things

1- Someone to compile a PERFECT source of exactly the codecs we need

2- A VERY simple VERY fast as close to lossless tool as possible with a SINGLE purpose. Allow me to chop up videos from my HD1

Join - Seperate and MOST importantly allow me to SPECIFY visually a start and end point and cut that out as a NEW clip with minimal recompressing none if possible and do this FAST.

THAT would just make my day.

As for #1 here is my issue. my desktop - reads perfectly. no problem playing no problem converting to divx with dr divx and no problem sicking TMPGENC on it to make DVD's out of them.

I tried it on my laptop and NOTHING but trouble. DrDivx gives me a GREY screen resultant file with audio (in preview it sees it just fine)

TMPGENC wont TOUCH them says file not supported. Clearly this is a codec issue (I tried every damned codec I could find) because TMPGENC at home see and interacts with these files JUST fine !!

I am thinking its time to refresh the laptop hopefully I can install the same stuff that I installed on the PC but I am not CERTAIN as to what I did different on the PC verse the Laptop (its a very powerful laptop more powerful than my desktop even so thats not it)

So if someone could do those 2 things I would be in heaven. I am REALLY liking this camera. If I can satisfactorily protect the camera (working on that right now) I am going to strap it to an RC car and see how that goes. its not a great car but it can do 18mph (well maybe not with this camera on its back for as small as it is its a heavy camera)

IF and ONLY IF I have enough money to buy another HD1 by august. If that happens I am going to put this camera in a Rocket. if it was half as heavy as it is I would not be so worried but as heavy as this camera is I am not confident I can pack it well enough to survive a fall from 5000+ feet. It will probabl go terminal before hitting the ground but terminal for this thing it likely over 200mph :-)

I figure I will need at least 12 inches of Foam Rubber padding in all directions to sufficiently protect it. I am also working on a rig to EJECT the camera from the rocket with its own attached parachute in case something goes horribly wrong. that will at least give me 3 chances to bail. Motor Ejection - Electronic Ejection and Emergency Camera Ejection. that should bring the odds of a good chute up considerable. at that rate I could get away with 3-6 inches of foam. The Trick will be protecting the most prone to damage part ie the lens since it has to stick outside the rocket :-) if I add too much foam I will trash the aerodynamics which will require more power which increases cost and risk. (its gonna take a hell of a motor to get this camera to 5k and my ultimate goal of 10k)

Chris Taylor
http://www.nerys.com/

It probably will be better to use in camera memory then SD card, because rewriting constantly to SD can damage the memory (though that might not be a problem now).

The compromise for this camera would be to continually compress and save to memory. This might consume more power, but you get many times more in the buffer. So 5 seconds, you could have 30 seconds or more.

Actually, if you treated the whole card as one big variable buffer then you could have many minutes, and avoid to many rewrites. Perfect for news people, they could be able to afford to throw away the odd wasted card every now and then.