Ask a Question
GoPro Support Hub Ask a question. Share an answer. Find a solution. Stay stoked.
Announcements
Is your GoPro gear up to date? Check to see If it is on our Update page.
Cameras
Reply
Nomad
Posts: 440

Re: HERO8 Firmware 1.60 - Robotic Jerky Pans

[ New ]

FAQ to Captering:

https://community.gopro.com/t5/en/GoPro-Camera-File-Chaptering-Information/ta-p/390210

It explains why ... But that's only in connection with the stabilization, which is the real reason for the complaint, just an unpleasant consequence ...

Hiker
Posts: 104

Re: HERO8 Firmware 1.60 - Robotic Jerky Pans

[ New ]

Yeah, I don't have any issues with chaptering, under normal cirmcumstances, other than it just being another step in my workflow that I would rather not be there, or something else that could break or go wrong. I've been joining chapters together for years. It's chapters plus ReelSteady demo that's caused me issues yesterday. I do think it's that 3 minute limit that's causing the lapse in time at the end of each chapter. If the Osmo Action doesn't meet my needs, I will probably just suck it up and buy ReelSteady and give that a go, even though it takes for dam ever to process a 1+ hour video. The ReelSteady does a fantastic job otherwise.

Hiker
Posts: 104

Re: HERO8 Firmware 1.60 - Robotic Jerky Pans

[ New ]

Osmo Action stabilization works perfectly for my needs after testing it. Plus, same FOV as HERO8, replaceable lens, 4K 60 Linear, 100mbit/sec, the mimo app lets my monitor my footage as I record at 4K 60 (which I can't do with GoPro, had to buy an external monitor and the media mod and hdmi cables). The cost of the camera that works right out of the box with the CYNOVA mic adapter and custom cage is less than 300.00 USD. 

 

If you do driving videos and want 4k 60 Linear, DO NOT buy the HERO8. Buy the OSMO Action. The stabilization does not work right for the HERO8 when shooting 60fps at 2.7 or 4 K.

 

I bought this too, freaking thing of beauty: https://www.smallrig.com/smallrig-cage-for-dji-osmo-action-compatible-with-microphone-adapter-cvd247...

Backpacker
Posts: 1,107

Re: HERO8 Firmware 1.60 - Robotic Jerky Pans

[ New ]

hi all,

 

here are my tests

  • the box test – if you are tldr, skip this is long, and the chopstick pan test has the same conclusion
    • how the touch screen and the actual footage look
    • 90-180 degree pan (includes up-down)
    • circular moves
  • the chopstick pan test
  • the chopstick outside test

all made in 4K 25fps Linear in all hypersmooth off, on, high, boost

 

to me all the results are normal, I see no issue.

 

the box test H8 - shows how much the gyro reading counts in the final result (and prove course if the gyro works) - if tldr skip it has the same results as  the chopstick pan tets

  • all touch screen looks the same in all hypersmooth both pan and circular movements (please take into account the increased crop)
    • the touch screen realign the centre with the same speed 
  • though the outcome is different and shows the degree the gyro reading is taken into account when the final footage
    • hypersmooth on
      • the gyro reading count the less
        • pan - the outcome is like the hypersmooth off
        • image remains centred
        • circular - almost centred
    • hypersmooth high and boost
      • the gyro count more and more
        • therefore the image centre slowly and with a delay remains more in the off position
        • image is behind the move
        • there is a slight difference between high and boost

the chopstick pan test - the same results as in the box test, but now the box is the chopstick 

  • pan in hypersmooth on
    • is the same as in hypersmooth off, gyro counts just a little
    • the chopstick remains in the centre and follows the pan
  • pan in hypersmooth high and boost
    • gyro readings make the chopstick to remain off centre long time after the pan move ends
    • it is like the image remains behind the pan
    • this could be an explanation for pan issue, but only if you use hypersmooth high and boost
    • the front of the car will mode ar the chopstick does in this test

the chopstick test H8

  • as I said before this is the real stabilization test 
  • the moves of the chopstick show the degree of movement compensated by the stabilization.
  • it is clear how the higher hypersmooth stabilize more.

 

conclusions if I may.

  • when panning in hypersmooth high and boost the image remain off centre, behind the pan movement with a delay
  • image remains centred and follow the pan move when hypersmooth on.
  • in the car the G forces are greater, so maybe the effect could be pushed to the limits, creating the robot pans when the image jump

I would not use hypersmooth at all when filming from the car.

But if you want to remove the vibrations, the hypersmooth on is enough and less prone to strange artefacts due to the image that remains behind the turn (the tip of the car moves in the pan direction as the chopstick does)

I started to let hypersmooth on (Superview) when driving, but I see no issue.

 

just me,

robert

 

 

https://youtu.be/51gVI2ZX1x0

https://youtu.be/n5aOnVFRr7k

https://youtu.be/c31-4TlGLh0

Hiker
Posts: 104

Re: HERO8 Firmware 1.60 - Robotic Jerky Pans

[ New ]

Hi Roboert,

 

I appreciate all of this data, however I could not reproduce the issue unless I was 2.7k or 4k 60fps in a car. If you want to reproduce the issue, try those settings. I replicated it on two different HERO8 cameras. However, be aware that when the issue happens, it only happens sometimes, as in for the entire recoding session or not for the entire recording session. For example I hit record and recorded over 1 hour of driving a few days ago and the issue didn't happen at all. Then I hit record the next day and the issue happened on every turn for the entire session. 

Hiker
Posts: 6

Re: HERO8 Firmware 1.60 - Robotic Jerky Pans

[ New ]

The problem is better described as 'car turns left, but camera pans right'. In contrast, humans turn their head into the turn: car turns left, head/eyes turns even more left to 'look ahead'. This is why it's so jarring, eg uncanny valley.

This is a result of 1) having the car in frame and 2) driving on a road. Both factors make it easy to see direction of movement.

 

This issue exists for all Hypersmoothed videos, but if the viewer can’t identify the direction of movement, they don’t realize this panning effect is going on, and its not as jarring.

 

Other scenarios that highlight this effect include: kayaks, boats (mounted on a pole).

 

Hypersmooth tries to minimize image motion frame to frame, and in doing so will dampen any angle changes (eg turning) experienced by the camera. While this works great on high frequency angle changes (aka vibration), it falls apart on long sweeping turns (eg driving).

 

This is a deficiency of how Hypersmooth is tuned.


@oliverz513 wrote:

Recording WITHOUT image stabilization and then stabilize with ReelSteady ....


ReelSteady has this exact same problem, it is not a silver bullet.

Hiker
Posts: 6

Re: HERO8 Firmware 1.60 - Robotic Jerky Pans

[ New ]

Instead of having just simple Off, High, Boost settings for Hypersmooth, GoPro should expose additional settings that allows us to tune Hypersmooth for our use case (eg for high frequency vibrations vs low frequency vibrations).

 

GoPro's one size fits all approach means if they fix it for one use case (eg vloggers), it breaks for other use cases (eg high motion but fixed/stable mounts, such as on a car).

For car mounted use (95% of my filming), I would like Hypersmooth to filter out vibrations (high frequency movement) but preserve turns (generally low frequency, and usually long sweeping, even in motorsports) and turn into the turn (car turns left 10°, camera pans left >10°, much like a human would). I think this could be accomplished by incorporating gyro data into Hypersmooth computation.

Hiker
Posts: 104

Re: HERO8 Firmware 1.60 - Robotic Jerky Pans

[ New ]

The OSMO Action does just that. It filters out the high frequency vibrations and leaves a lot of the other movements. I agree that GoPro needs to address this and I will consider using their cameras again. 

 

For me, the ReelSteady doesn't have the pan issues but makes the car seem more like a drone that's floating across a surface of glass.