What is best practice for getting a video file's codec?

Wishlist and RFEs
1

with proRES finally supported in experimental (YAY!!!) I’m finding myself in a position of needing to determine the codec of media I’m pointing at quickly; is there a supported best practice for how to best determine if something is HAP, prores, h264, DXV, notchLC etc?

perhaps making this a member of the moviefileinTOP would be the most TD approach to dependably being able to determine codec info?

Thanks!

2

I use this:

3

Yes, there are a million ways to get this data outside of TD; I want access to this information inside of TD

4

Then you did not phrase your original post correctly. This is a RFE for TD to provide mechanisms to extract metadata out of the supported movie files, which it does not provide.

You can, however, use the CLI executable of MediaInfo to programatically poll and get this information into TD, until it is supported natively.

Edit: To expand on this, I think supporting metadata would be ideal as part as the folder DAT, allowing one to select which columns to display, such as codec, container, resolution, frame rate, etc…

5

Look out for the new Media Info DAT in the next Experimental - let’s see if it extracts what you need.

2 Likes
6

Love that!

any chance it can be used to extract or add meta data to the file?

Touchdesigner has become a great way to control local generative AI image generation through Stable Diffusion APIs like ComfyUI or Automatic1111, and when you’re generating images with lots and lots of parametric data associated with their prompts, seeds, samplers, models etc., being able to add arbitrary meta information to the result itself makes things way more portable than a sidecar JSON or other solution to the problem of associating this data.

I know TD currently can write data to EXRs, but most of the supported media types in TD can work with some format of meta data, and if this Media Info DAT had a method of accessing and updating the meta data of a file that would be huge. ComfyUI even implements this by storing all of the data of the network that generated an image in the meta data, the equivalent of storing a .toe in .png meta text, which makes dropping an image file into the network a super portable way of recalling a state, a clever image result with all of the ingredients incorporated