Thanks Dave. Glad to hear its not my drives and that its all working properly. Its not really an inconvenience and seams favorable when you take into account the possible corrupt file issue.
HDMI Atomos Ninja Hard Disk Recorder 795$
- Thread starter headroom
- Start date
Thanks Dave. Glad to hear its not my drives and that its all working properly. Its not really an inconvenience and seams favorable when you take into account the possible corrupt file issue.
Shoot2Thrill
New member
look my goal is to use 2 gh2 cameras and have both running ninjas so I can use lanc to control them for 3d
I no they also make another model with sdi but that means adding two other devices to the cameras
regardless if the ninja does work with gh2 cameras I want 2 of than as pro red is far more user friendly than avchd
and makes the editing far more simple
I no they also make another model with sdi but that means adding two other devices to the cameras
regardless if the ninja does work with gh2 cameras I want 2 of than as pro red is far more user friendly than avchd
and makes the editing far more simple
Shoot2Thrill
New member
the thread says ninja 795$ well if I can get them for that price I want 2 for sure
Jarek Zabczynski
Veteran
SO EFFING TRUE!!! I've never had a signal cut out so much on any other piece of equipment I've ever owned.
David Shapton
Well-known member
Jarek - I don't doubt this, but I just wanted to say that with proper cable management, this needn't be a problem. The Ninja's HDMI port is very tight and secure. We have enough of them out there to know that this isn't an issue.
Dave Shapton
Atomos
David, we recently purchased the AF100/102 and are thrilled at the results. We also purchased the Ninja to take advantage of the clean hdmi out which the AF100 provides. I read in this thread you said taking samples takes time and you're sorry. So over the last 3 days we took it upon ourselves to test out the footage for you. Comparing the AVCHD footage (transcoded to ProRes HQ in Final Cut Pro 7) to the Ninja (AF100-HDMI out to ProRes HQ) footage.
After rigorous testing we came to the conclusion that they were identical, in quality and durability.
Both footages in all scenarios from low light, outdoors, close ups, colour charts, green screen and focus charts have the same results under colour correction, gamma lifts, keying footage, over cranking, under cranking and flesh tones. Pushing the footage to its full extent to find exactly where the 1000 dollar difference was by recording via hdmi. We will soon publish a full report on our findings.
After rigorous testing we came to the conclusion that they were identical, in quality and durability.
Both footages in all scenarios from low light, outdoors, close ups, colour charts, green screen and focus charts have the same results under colour correction, gamma lifts, keying footage, over cranking, under cranking and flesh tones. Pushing the footage to its full extent to find exactly where the 1000 dollar difference was by recording via hdmi. We will soon publish a full report on our findings.
David Shapton
Well-known member
@Capcap - I'm glad you found them to be visually identical because it confirms that HDMI is a valid way to record for extra record times and low-cost storage. As far as quality is concerned I think it is important to understand what we are claiming (in fact, not us, but Apple, because it's their codec). Apple claims that ProRes is visually lossless - in other words, that it is visually identical to uncompressed. So what you are effectively saying is that if your recordings are identical to the ProRes ones, then AVCHD is visually identical to Uncompressed. That's certainly a bold claim.
Quite obviously, the output from your camera is very good. If it wasn't then it probably wouldn't be selling very well. In these circumstances, then ProRes is NOT going to make a magical degree of improvement to your footage. In fact, if it did, we would be charging massively more for it, because you would be getting the equivalent of a camera that costs perhaps ten times as much for $1000. You have to pay a lot for a small improvement at this level.
No, for high bitrate AVCHD footage that has been fed from a good sensor, you probably won't notice much - if any - difference visually. But when you take more challenging images, with sharp edges and with complex backgrounds (foliage, crowds) and with swirling motion (like water) then you will see the undeniable advantage of an I Frame-only codec over a Long-GOP one. By concentrating on charts and flesh-tones, you're not entering the difficult areas for AVCHD.
There's no doubt that if you push a Long-GOP codec to a high enough bitrate it can certainly look very good. And this is also why I'm disinclined to publish comparisons on a website, because it's missing the point.
Look, if you're saying that we don't give $1000 worth of quality improvement, you've misunderstood what we're selling with the Ninja.
You get a monitor, a hardware recorder, a playback device, a touchscreen, a case, batteries and a charger. For $1,000. That's what we're selling. On top of that, you get a professional production codec that's easy on processors and which doesn't suffer from motion artifacts. We've seen cases where there is a distinct improvement in quality, and we've seen some where our encoded footage is merely visually identical to the original. It is definitely the case that ProRes HQ is a better codec visually than AVCHD, but AVCHD is itself a remarkably capable codec at higher bitrates, and sometimes, you'll be hard-pushed to tell the difference, under favorable conditions.
By the way, if you were comparing AVCHD to ProRes by first converting ACVCHD to ProRes, then all of your grading and gama lifts would have been done at 10-bit, even though the original material would have been 8 bit.
I'm not going to get drawn into this. ProRes is one of the codecs we've chosen to support, because it was the one most requested by potential customers. It gives great results consistently and particularly in those cases where Long GOP shows weaknesses.
Part of the value of the Ninja is that it can run any codec that we choose. It could even run AVCHD if we decided to code it. (We won't be doing that, by the way!).
Dave Shapton
President
Atomos EMEA
Quite obviously, the output from your camera is very good. If it wasn't then it probably wouldn't be selling very well. In these circumstances, then ProRes is NOT going to make a magical degree of improvement to your footage. In fact, if it did, we would be charging massively more for it, because you would be getting the equivalent of a camera that costs perhaps ten times as much for $1000. You have to pay a lot for a small improvement at this level.
No, for high bitrate AVCHD footage that has been fed from a good sensor, you probably won't notice much - if any - difference visually. But when you take more challenging images, with sharp edges and with complex backgrounds (foliage, crowds) and with swirling motion (like water) then you will see the undeniable advantage of an I Frame-only codec over a Long-GOP one. By concentrating on charts and flesh-tones, you're not entering the difficult areas for AVCHD.
There's no doubt that if you push a Long-GOP codec to a high enough bitrate it can certainly look very good. And this is also why I'm disinclined to publish comparisons on a website, because it's missing the point.
Look, if you're saying that we don't give $1000 worth of quality improvement, you've misunderstood what we're selling with the Ninja.
You get a monitor, a hardware recorder, a playback device, a touchscreen, a case, batteries and a charger. For $1,000. That's what we're selling. On top of that, you get a professional production codec that's easy on processors and which doesn't suffer from motion artifacts. We've seen cases where there is a distinct improvement in quality, and we've seen some where our encoded footage is merely visually identical to the original. It is definitely the case that ProRes HQ is a better codec visually than AVCHD, but AVCHD is itself a remarkably capable codec at higher bitrates, and sometimes, you'll be hard-pushed to tell the difference, under favorable conditions.
By the way, if you were comparing AVCHD to ProRes by first converting ACVCHD to ProRes, then all of your grading and gama lifts would have been done at 10-bit, even though the original material would have been 8 bit.
I'm not going to get drawn into this. ProRes is one of the codecs we've chosen to support, because it was the one most requested by potential customers. It gives great results consistently and particularly in those cases where Long GOP shows weaknesses.
Part of the value of the Ninja is that it can run any codec that we choose. It could even run AVCHD if we decided to code it. (We won't be doing that, by the way!).
Dave Shapton
President
Atomos EMEA
Last edited:
Alvise Tedesco
Well-known member
I always enjoy test and reviews. But with a first post like this, better if you give links to your activity instead of announcing a future published report
BobbyMurcerFan
Well-known member
I too am very skeptical of this. I've seen many examples of nanoFlash footage improving Sony EX-1/3 footage. Going from 4:2:0 to 4:2:2 alone makes the recorder footage better.
I also think that's a lot of testing:
1) Low light
2) Outdoors
3) Close ups
4) Colour charts
5) Flesh tones
6) Green screen
7) Focus charts
8) Over cranking
9) Under cranking
And evaluate these nine types of shots under:
A) Colour correction
B) Gamma changes
C) Keying
That would mean doing a minimum of 54 comparisions (9 * 3 * 2 = 54).
Barry_Green
Moderator
It depends on what was shot. In some cases they probably do look pretty close to even identical; look back on the Crews.TV report that the AF100's AVCHD looked almost indistinguishable from AVC-Intra(!) Depending on the type of shot, if AVCHD can record it well, then there's not that much more for a better codec to do.
But, as David said, there are cases which will tax or exceed the low-bandwidth long-GoP codec, and that's where you'll start seeing the difference. On an interview? No problem, and I do pretty much dare anyone to be able to tell the difference. But on an action scene doing some Bourne-style stuff in a daylight forest? There will be a substantial, if not huge, difference between the two.
And, in all cases, you're going to be looking at 4:2:2 vs. 4:2:0. 4:2:2 is just better, no matter how you slice it. If you can't see the difference, well, I dunno, because you should be able to; there's a reason the BBC requires 4:2:2 acquisition in any camera they approve for production. But for greenscreening or compositing or such, 4:2:2 does make a substantial difference.
Unquestionably, the Ninja's codec is and should be superior to AVCHD. It's 10x the bandwidth, it's intraframe, it's 4:2:2 for twice the color, and it's 10-bit (for cameras that can take advantage of that). Under certain scenarios that superiority is simply going to be obvious, but under other scenarios the AVCHD will perform well enough that yes, it's reasonable for people to think that they look very comparable. It's like this -- when the DSLRs first hit, we had lots of people comparing them to Reds and saying "look, they look just as good!" And in certain highly controlled scenarios, they did look really good, especially when viewed on vimeo. But the idea that a DSLR, overall, in all cases, produces footage as high quality as a Red, especially in any scenario other than a head shot, is laughable. And I think over the intervening years, as people have worked more with the DSLRs, they now understand that. And the SCCE clearly exposed the gulf of quality difference between them. Same thing will apply here: unquestionably a high-quality external recorder is capable of producing better results than the onboard 24mbps 4:2:0 long-GoP codec. The question is -- will you be encountering those scenarios that exceed the onboard's capability? If not, then no, you won't see the difference. But if you do exceed it, you'll clearly see the differences.
And, finally, you'll have to decide whether the difference is worth the cost. For some it's a no-brainer, of course it is. For others it's a no-brainer, of course it isn't. That's an individual decision you'll have to make.
But, as David said, there are cases which will tax or exceed the low-bandwidth long-GoP codec, and that's where you'll start seeing the difference. On an interview? No problem, and I do pretty much dare anyone to be able to tell the difference. But on an action scene doing some Bourne-style stuff in a daylight forest? There will be a substantial, if not huge, difference between the two.
And, in all cases, you're going to be looking at 4:2:2 vs. 4:2:0. 4:2:2 is just better, no matter how you slice it. If you can't see the difference, well, I dunno, because you should be able to; there's a reason the BBC requires 4:2:2 acquisition in any camera they approve for production. But for greenscreening or compositing or such, 4:2:2 does make a substantial difference.
Unquestionably, the Ninja's codec is and should be superior to AVCHD. It's 10x the bandwidth, it's intraframe, it's 4:2:2 for twice the color, and it's 10-bit (for cameras that can take advantage of that). Under certain scenarios that superiority is simply going to be obvious, but under other scenarios the AVCHD will perform well enough that yes, it's reasonable for people to think that they look very comparable. It's like this -- when the DSLRs first hit, we had lots of people comparing them to Reds and saying "look, they look just as good!" And in certain highly controlled scenarios, they did look really good, especially when viewed on vimeo. But the idea that a DSLR, overall, in all cases, produces footage as high quality as a Red, especially in any scenario other than a head shot, is laughable. And I think over the intervening years, as people have worked more with the DSLRs, they now understand that. And the SCCE clearly exposed the gulf of quality difference between them. Same thing will apply here: unquestionably a high-quality external recorder is capable of producing better results than the onboard 24mbps 4:2:0 long-GoP codec. The question is -- will you be encountering those scenarios that exceed the onboard's capability? If not, then no, you won't see the difference. But if you do exceed it, you'll clearly see the differences.
And, finally, you'll have to decide whether the difference is worth the cost. For some it's a no-brainer, of course it is. For others it's a no-brainer, of course it isn't. That's an individual decision you'll have to make.
dustylense
Veteran
I think Barry nails it right there.
I bought the Ki Pro Mini and love it. I get the 4.2.2 bonus plus I record on the SD card and Ki Pro at the same time, getting a good backup of my files. Drag and drop editing a plus as well. But the biggest thing I see that you are capturing for requirements the likes of the BBC and such.
Is there a BIG difference in the external recorder (Ki pro mini in my case). Not really. I think the AVCHD holds pretty well on it's own. But after testing and playing around, the Ki pro footage is better. $2,500 better? (the amount i paid for recorder, mounting and battery). In shots that have lots of detail, I will trust the recorder over the onboard codec. I think it's nice to have both. So which recorder you go with is up to you. I still have not seen very many "good" examples of the Ninja, which is why I went Aja....
I bought the Ki Pro Mini and love it. I get the 4.2.2 bonus plus I record on the SD card and Ki Pro at the same time, getting a good backup of my files. Drag and drop editing a plus as well. But the biggest thing I see that you are capturing for requirements the likes of the BBC and such.
Is there a BIG difference in the external recorder (Ki pro mini in my case). Not really. I think the AVCHD holds pretty well on it's own. But after testing and playing around, the Ki pro footage is better. $2,500 better? (the amount i paid for recorder, mounting and battery). In shots that have lots of detail, I will trust the recorder over the onboard codec. I think it's nice to have both. So which recorder you go with is up to you. I still have not seen very many "good" examples of the Ninja, which is why I went Aja....
David Shapton
Well-known member
Not sure I understand that. Ninja ProRes is ProRes. Apple makes sure of that, believe me!
Dave Shapton
Atomos
hemophilia
Well-known member
Forgive me if some has been posted, but this thread is getting pretty long. Has anyone shot some A/B out there? I think that's all dustylense is saying. I too would love to see some comparisons to quantify a bit what this improvement is like. I agree with everything everyone in the past few posts has said, notably Barry. But I would much prefer to see with my own eyes what sort of difference people are getting.
dustylense
Veteran
That's all I've asked for a while now. Just some samples from the device. Hell, even the Atomos website does not have any samples from the Ninja. But they say prores is prores. OK cool. But a device is a device? I like what Atomos has developed and at it's price. But samples help people understand if they justify buying it. REAL WORLD SAMPLES.
Here is the only justifiable A/B footage I've seen from the Ninja to date.
http://vimeo.com/24706136
I think it would be unwise to conflate the performance of the Ninja with that of the AF100.
Don't forget that the Ninja can be used with many HDMI capable cameras, including HDV and XDCAM models, where improvements between sensor output and the capabilities of onboard compression will be more marked.
Don't forget that the Ninja can be used with many HDMI capable cameras, including HDV and XDCAM models, where improvements between sensor output and the capabilities of onboard compression will be more marked.
David Shapton
Well-known member
It's great that people are doing these comparisons. The vimeo clip above isn't going to stress any codec, so it won't highlight the differences between AVCHD (a long-GOP codec) and ProRes (an "I-Frame Only" codec).
There's a bigger issue with these comparisons on Vimeo, though, which is that everything is transcoded into a Long-GOP format. Which means that you don't know whether you're seeing artifacts from the original AVCHD, ProRes or from the H.264 codec used by Vimeo. It's exactly the same as doing a comparison between two prime lenses through a transparent plastic bag.
Dave Shapton
Atomos
http://vimeo.com/24706136
cheezweezl
Veteran
I recently did a quick and dirty test, comparing onboard AVCHD to Prores via the HDMI output. I haven't decided which recorder I am going to buy so I did this test by recording Prores straight into the Mac via a blackmagic card. So unless I'm missing something, this should be identical to using a ninja or some other prores device. I knew I wouldn't really see much difference unless the codec was being taxed so I shot in low light, waving my hand in front of the lens. I found a huge difference.
See here..
http://www.dvxuser.com/V6/showthread.php?251432-AVCHD-in-cam-vs.-Prores-HQ-capture-via-HDMI
See here..
http://www.dvxuser.com/V6/showthread.php?251432-AVCHD-in-cam-vs.-Prores-HQ-capture-via-HDMI
colinhubick
Member
The good and bad
The good and bad
I bought one of the Atomos Ninja's about 3 weeks ago, and have had quite a few instances where we used it. As with many "low-cost" solutions like this, they don't always live up to the expectation you have for them.
Here are some of the positive attributes of the Ninja
- low cost of unit, good value for your money (kit does not include media like many other recorders)
- cost of media, and availability of media (extremely low cost compared to other media solutions, and readily available)
- form factor (the form factor is very good, very compact and does not add to much weight)
- image quality (the quality of the encoded prores files is exactly what you would expect from prores, nothing to worry about here.)
Here are some of the negative attributes of the Ninja
- playback on the current firmware 1.06 is terrible. Unusable.
- if you are shooting 24p, like many of us are, you will need to reverse telecine the footage from the Ninja to get the footage back to it's native 23.976 frame rate, to be working in true progressive frames. This currently needs to be done via software on your system, so if shooting 24p, and want to work at 23.976 with progressive frames, you should allocate time to do this.
- all clips are recorded to the ninja with the same file name, 000.mov. I have no idea why this is the case, and from a post production workflow point of view could be dangerous. Imagine trying to relink a project where all clips have the same name. Not fun. To add further issues on this side of things, the current firmware 1.06 does not use TOD timecode. So at the end of the day you have a drive with clips that all have the same name and all start at 00:00:00:00. According to Atomos, there will be a firmware out soon that will add TOD timecode, but have said nothing about the duplicated clip names.
Just my experience with this unit so far.
Would love to know when the firmware is going to be ready that add's TOD, and fixes the playback…
cheers
Colin
The good and bad
I bought one of the Atomos Ninja's about 3 weeks ago, and have had quite a few instances where we used it. As with many "low-cost" solutions like this, they don't always live up to the expectation you have for them.
Here are some of the positive attributes of the Ninja
- low cost of unit, good value for your money (kit does not include media like many other recorders)
- cost of media, and availability of media (extremely low cost compared to other media solutions, and readily available)
- form factor (the form factor is very good, very compact and does not add to much weight)
- image quality (the quality of the encoded prores files is exactly what you would expect from prores, nothing to worry about here.)
Here are some of the negative attributes of the Ninja
- playback on the current firmware 1.06 is terrible. Unusable.
- if you are shooting 24p, like many of us are, you will need to reverse telecine the footage from the Ninja to get the footage back to it's native 23.976 frame rate, to be working in true progressive frames. This currently needs to be done via software on your system, so if shooting 24p, and want to work at 23.976 with progressive frames, you should allocate time to do this.
- all clips are recorded to the ninja with the same file name, 000.mov. I have no idea why this is the case, and from a post production workflow point of view could be dangerous. Imagine trying to relink a project where all clips have the same name. Not fun. To add further issues on this side of things, the current firmware 1.06 does not use TOD timecode. So at the end of the day you have a drive with clips that all have the same name and all start at 00:00:00:00. According to Atomos, there will be a firmware out soon that will add TOD timecode, but have said nothing about the duplicated clip names.
Just my experience with this unit so far.
Would love to know when the firmware is going to be ready that add's TOD, and fixes the playback…
cheers
Colin
David Shapton
Well-known member
Hi Colin
Thanks for buying a Ninja and I'm glad you're seeing the benefits from it.
Let's deal with those negatives.
- Playback.
Yes, completely agree. We're very close to releasing an upgrade that will do playback in hardware (perfect) as opposed to software (er...). This is a major upgrade and it's taken way longer than we'd hoped or expected. We're very sorry about that to all Ninja users. All existing and future Ninjas are fully upgradable to the new playback capabilities.
- 24p.
Yes, this is a fact of life with HDMI recorders. The Ninja DOES record 24p over HDMI, and if only the manufacturers would supply a 24p signal over the HDMI cable, then we'd all be very happy. Meanwhile, we're finishing off "Stripper", which is our rapid pulldown removal software. (I haven't seen it running yet so I can't say exactly how fast!).
- Clip naming.
Not ideal. We're changing it. The first thing we're doing is implementing Time Of Day timecode, and this will be in the same imminent release as the Hardware Playback. It's very close now.
Dave Shapton
President
Atomos EMEA
Thanks for buying a Ninja and I'm glad you're seeing the benefits from it.
Let's deal with those negatives.
- Playback.
Yes, completely agree. We're very close to releasing an upgrade that will do playback in hardware (perfect) as opposed to software (er...). This is a major upgrade and it's taken way longer than we'd hoped or expected. We're very sorry about that to all Ninja users. All existing and future Ninjas are fully upgradable to the new playback capabilities.
- 24p.
Yes, this is a fact of life with HDMI recorders. The Ninja DOES record 24p over HDMI, and if only the manufacturers would supply a 24p signal over the HDMI cable, then we'd all be very happy. Meanwhile, we're finishing off "Stripper", which is our rapid pulldown removal software. (I haven't seen it running yet so I can't say exactly how fast!).
- Clip naming.
Not ideal. We're changing it. The first thing we're doing is implementing Time Of Day timecode, and this will be in the same imminent release as the Hardware Playback. It's very close now.
Dave Shapton
President
Atomos EMEA
Yeah not to nitpick, but this footage has image noise. The ninja will not clean noise thats already produced in the sensor. How can it? It can only record what comes through the sensor.
Bottom line, shooting an underexposed shot to create a lot of noise and then comparing that noise on an AVCHD recording vs that noise on an external recorder to extrapolate the difference in recording quality is an invalid test. You will not a few have mention high motion shots being the place to notice the difference and they do so because its the valid comparison.
I have a ninja in my hands that I need to do some testing on before I head to Japan next week - one test I will do is a lake by my house. In a reasonable wind, and a wide shot this fairly shallow lake has a huge amount of fine detail surface motion. Its the only thing I have come across that has truly taxed the AVCHD codec to the point where Id say its reached its limit (fwiw I wouldnt generally shoot a shallow lake in a high wind, well I cant ever recall having had a need to). Ill put a side by side up where Im pretty the difference is going to be clearly evident.