Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Any idea when the next firmware will come out of beta for Komodo?

Here's the test I promised with RED Control app ver.1.9.6. My own impression is that, if anything, the autofocus has actually gotten worse. I was going to do yet another test this morning, but the camera just refused to track at all and I gave up. Again, it's best watched on an HDR capable device.

 
Not a great performance, but I am now wondering if it does better if the face to be tracked stays near the center of the frame. That is, suppose the camera was moved to follow you around instead of just static. Would that have made a difference?

It seems unlikely that RED would claim aufofocus improvements and not deliver them, but what do I know?

Oh, and Merry Christmas to all!

Thanks.

Stephen
 
This is of course not the big AF update Jarred Land mentioned earlier that uses AI and is built into the camera.
Also not sure how dependent the face tracking in RED Control is on any firmware updates from RED since it seems
to be functionality that is based more on the app.

Would be good though if RED specified the areas of improvements with AF in tis latest update.

The current AF with face detection has already been helpful for a couple of specific shots I've had to do but you definitely have to
work within its limits.

Perhaps RED and Mikael from Fool Control might consider working with Jon as a beta tester for the AF developments.
I'm sure this is going to be an ongoing development that takes various iterations to get right.

Brian Timmons
BRITIM/MEDIA
​​
 
As I understand it, these latest AF improvements are for the camera firmware and therefore not so relevant for face tracking, which is still mostly based on the algo in the device, not the app. In the scenario where the person is always foreground, the in-camera AF might give better results than face detection AF in the app. Face detection AF will probably benefit from the new firmware in that it nails the focus better in the place that face detection points it to, but the place where the face is found in the first step only depends on where the device finds it.

For face detection, the biggest issues for me are that the focus, once it temporarily loses the face, switches to the background in the place where the face has last been found which often happens when the face moves horizontally across faster. The app could probably handle this situation and suppress focus adjustments when the face is lost only for a few seconds. Also, the inability to detect a face when it doesn't look straight to the lens. Look to the side and it's gone most of the time. Or more generally, when not both eyes are visible, it's often no longer recognized. Just try temporarily covering one eye with your hand and focus gets lost when you move a bit like that. A bright light from behind the head can also distract it but for that scenario I'd expect reliable results only with lidar anyway.
 
Here's a first look at HQ and ELQ frame grabs, enlarged 999%. :))

https://daejeonchronicles.com/2022/1...hq-vs-r3d-elq/

To really evaluate the image quality and see what compression is doing to our footage, it's necessary to play the clips back-to-back in HDR on something like a 55" LG OLED, and when doing so, I don't find the quality acceptable even for YouTube. I may post a video or share a downloadable R3D file later this week.
 
Last edited:
I cannot help but wonder how I was able to live without Gyoscope before. It was one of these features that I did not care much about until I tried it. Setting camera display to Gyo and external display to LUT view is just a pleasure to shoot with. Wow!
 
Jon Pais said:
Click to expand...

Jon- Congratulations for being one of the 1st people to actually upload a test showing ELQ compared to another flavor of REDcode.
Despite ELQ being available (albeit on only V-Raptor) since this summer I have not seen any tests uploaded.

I'm tied up editing a project and have not been able to test it extensively (not to mention it's brick cold where I'm at) but I hope to spend a day filming in ELQ to understand the visual limits.
Granted Youtube's own compression is not a completely accurate picture of how good it is but considering that Youtube is a popular delivery platform it's good to see how it behaves there.

Based on the quick shots I have done with ELQ I think it's pretty decent considering it's close to 14:1 compression. TBH I was expecting a bit worse macro blocking for the DCT compression that's now used in DSMC 3's version of REDcode.

I imagine the macro blocking will be more pronounced at 4K and under and that level of compression likely works better with higher resolutions.

I still think it's great that RED at least put it out there as an option.

There is definitely room between LQ and ELQ for another flavor (perhaps 10:1) of REDcode to be developed but ELQ goes a decent way in covering the bases.

Thanks again for the test.

Brian Timmons
BRITIM/MEDIA
 
Ian M said:
I cannot help but wonder how I was able to live without Gyoscope before. It was one of these features that I did not care much about until I tried it. Setting camera display to Gyo and external display to LUT view is just a pleasure to shoot with. Wow!
Click to expand...

I imagine you mean Gioscope the tool for evaluating exposure and not the gyro data which is used to record camera positioning to be used for stabilization etc.
Yes- it's great that both are in the new firmware and they are quite helpful.
Shows that RED is listening to user requests.
I appreciate the work the RED firmware team did in getting this done before the end of 2022.


Hope 2023 is the year DSMC 3's feature set truly matures.
*Wireless video enchancements (more in the area of stability as it can still be a bit finicky).
*More resolutions and associated frame rates (3K modes at corresponding over cranked speed, 5K 16:9)
*Slower shutter speeds for better timelapse
*AI autofocus as previous hinted
*Faster Black shade in Komodo like what was developed in V-Raptor.
*an in-between REDcode compression between LQ and ELQ
*built in horizon level.


All of the above would be great and hopefully on the horizon but I understand these things take time and I'm willing to wait for RED to do what they have to do to get things right.

Until then, the current firmware update is quite good so far and has some tools that truly help get more out of the camera.

Brian Timmons
BRITIM/MEDIA
 
Brian Timmons said:
Jon- Congratulations for being one of the 1st people to actually upload a test showing ELQ compared to another flavor of REDcode.
Despite ELQ being available (albeit on only V-Raptor) since this summer I have not seen any tests uploaded.

I'm tied up editing a project and have not been able to test it extensively (not to mention it's brick cold where I'm at) but I hope to spend a day filming in ELQ to understand the visual limits.
Granted Youtube's own compression is not a completely accurate picture of how good it is but considering that Youtube is a popular delivery platform it's good to see how it behaves there.

Based on the quick shots I have done with ELQ I think it's pretty decent considering it's close to 14:1 compression. TBH I was expecting a bit worse macro blocking for the DCT compression that's now used in DSMC 3's version of REDcode.

I imagine the macro blocking will be more pronounced at 4K and under and that level of compression likely works better with higher resolutions.

I still think it's great that RED at least put it out there as an option.

There is definitely room between LQ and ELQ for another flavor (perhaps 10:1) of REDcode to be developed but ELQ goes a decent way in covering the bases.

Thanks again for the test.

Brian Timmons
BRITIM/MEDIA
Click to expand...
As in my previous comparison between HQ and LQ, I was only able to see a difference when watching on a 55” display. At a normal viewing distance, I was able to see a slight coarsening of the image which was borne out by the image comparisons I shared. With ELQ, I was again unable to notice much difference unless pressing my nose up against the display, where, if you look at say, the folds of the curtains just behind my neck, you can see the image beginning to fall apart. On both my MacBook and on the LG OLED, there appears to be a very slight change in tonality, too, but as I was shooting by available light, I can’t draw any conclusions.
 
Since updating the RED Control app, it seems as though buttons assigned to things like Auto White Balance disappear from time to time and I have to re-assign them, unless I’m losing my mind, which is another possibility.
 
Jon Pais said:
Click to expand...

Great test! ELQ doesn't seem bad in a pinch. LQ and above is still going to be the main codecs for the majority of work but it's always good to have options.
Be interesting to see how well it cleans up in noise reduction.

Brian Timmons
BRITIM/MEDIA
 
Alex Zeder said:
As I understand it, these latest AF improvements are for the camera firmware and therefore not so relevant for face tracking, which is still mostly based on the algo in the device, not the app.
Click to expand...

I'm getting better basic results (non face tracking) with the Canon 35mm F2 IS, before the update it would mostly search around before finding focus, now it locks on right away.
 
Jon Pais said:
Since updating the RED Control app, it seems as though buttons assigned to things like Auto White Balance disappear from time to time and I have to re-assign them, unless I’m losing my mind, which is another possibility.
Click to expand...

this custom button issue sometimes missing should now be fixed in 1.9.7 thanks.

FWIW I’m still testing different stuff but the 1.9.6 release was strictly identical to how it worked before as far as face detect / AF goes.
 
Mikael Lubtchansky said:
this custom button issue sometimes missing should now be fixed in 1.9.7 thanks.

FWIW I’m still testing different stuff but the 1.9.6 release was strictly identical to how it worked before as far as face detect / AF goes.
Click to expand...

Thank you so much for all the work you’ve put in, Mikael!
 
Is the accelerometer and gryo data embedded by default with this build or does it need to be enabled in settings somewhere? And is there a workflow to extract it for use in something like gyroflow, which i assume still can't open r3d natively?
 
Bobby Gupton said:
Is the accelerometer and gryo data embedded by default with this build or does it need to be enabled in settings somewhere? And is there a workflow to extract it for use in something like gyroflow, which i assume still can't open r3d natively?
Click to expand...
It’s embedded in the files, no need to do anything. It can be used with gyroflow, but footage needs to be transcoded to ProRes, use as video source in GF, use R3D as gyro source, export gyroflow file from GF, load directly into Resolve Gyroflow OFX applied directly to R3D video and you’ve just stabilized your RAW footage.
 
Will DaVinci resolve support Gyro stabilisation without any 3rd party plugins for R3D? Anyone has heard anything about it so far?
 
You must log in or register to reply here.
Back
Top