@g3gg0 updated mlv_dump and put up binaries for Windows, Mac and Linux so I removed the mlv_dump builds I was hosting. These new builds work with 10bit/12bit files.
http://www.magiclantern.fm/modules/modules/mlv_dump.zip/mlv_dump.zipSome of you are posting links to the test builds which is fine but just a warning that I've been removing old builds to avoid confusion. That means that you'll need to save previous versions in case something breaks or you want to make a regression test.
There have been some reports that things aren't working properly on a few cameras that should be working. It might be the way I'm compiling. In most cases I'm just compiling the raw_video_10bit_12bit branch but in others I merged crop_rec and other branches. Maybe I should be merging unified too since eventually the goal is to get this into the nightly builds?
Here are the latest download counts of the test builds. Checkmarks are for cameras that are working with CONFIG_EDMAC_RAW_SLURP:
5D3.123 | - | 76 | √ | * with crop_rec (Dec 20) |
5D3.113 | - | 55 | √ | * without crop_rec (Dec 18) |
5D3.113 | - | 51 | √ | * with crop_rec (Dec 24) |
6D.116 | - | 24 | √ | * new build works with 10bit/12bit (Dec 23) |
5D2.212 | - | 21 |
7D.203 | - | 15 |
EOSM.202 | - | 14 | √ |
100D.101 | - | 14 | √ | * might have build issues? (Dec 23) |
600D.102 | - | 13 | √ |
70D.112 | - | 13 | √ | * might have build issues? (Dec 23) |
700D.114 | - | 12 | √ |
60D.111 | - | 10 | √ |
650D.104 | - | 10 | √ |
1100D.105 | - | 9 | √ | * allocate_raw_buffer branch has CONFIG_EDMAC_RAW_SLURP working but not 10bit/12bit (Dec 19) |
550D.109 | - | 8 |
50D.109 | - | 6 |
500D.111 | - | 3 |
Although this might not be an accurate list of cameras sorted by popularity, it does show quite a bit of interest in 10bit/12bit raw video from users of the higher end and full frame cameras. I was surprised by the amount of interest in the 5D3.123--looks like users want raw video
and clean HDMI output.
There a few older models left to solve. Check out reply #754 for some hints:
From what I've tried so far on those old models (without much success yet), we have at least two problems:
- the raw buffer is not allocated all the time (as it's done on the new models) -> easy to fix, see the 1100D raw branch
- the vsync hook currently used doesn't appear to be the best for getting a raw image
- the exact behavior of EDMAC in case of input/output dimension mismatch is not very well known
Solving those models may require either some serious low-level research, or a great dose of luck. I'm just scratching the surface here.