So, after doing the User Liberation video for the Free Software Foundation campaign, I did a longish postmortem that you can read here. There's a little section in it about (video) compression that got me thinking: There's roughly two types of transcoding you can easily do in free software, from a user perspective:

• Command line stanzas, either using ffmpeg directly, or something like mencoder or gstreamer
• A user centric tool like Transmageddon or handbrake, that offers necessarily limited options and media type choices, based on presets for specific devices or outputs
• (I said roughly) built in (also usually preset based) encoding options in bigger programs like kdenlive, Pitivi, Blender, that usually expose a few under-the-hood knobs.

Ignoring bugs, this works OKish - but it is non-ideal for my specific use-case, lets say: animator / filmmaker who needs to do 'professional' mastering, produce multiple output types and sizes/ target media, and is not an expert on arcane options in video codecs.

• Command line gives you the ability to massively shoot yourself in the foot. we mainly end up cribbing ffmpeg stanzas off each other or the 'net - and these tend to break from time to time, as either the requirements change or ffmpeg does. In general, this tends to give videos that play on your own machine fine, but might break on different OS's / devices.
• Tools like transmageddon or handbrake may or may not have options you need, and feel fairly rigid: if you mess with the preset (Assuming you can) you might break the output. I still get non-working on all platforms video (especially using h264) but less so, and I can't work with my 'real' masters: image sequences and audio tracks - I have to start from a video, encoded losslessly - this means I have to do a pre-encoding to a lossless format before I can actually encode the video.
• Output from Blender is often quite limited, Pitivi is better, but I do get into the 'not enough feedback' issue: I don't really know what a knob does unless I output the video, and even then, sometimes it is a mystery what it did.

So what might be cool ideas / experiments?

• GUI tool for professionals: assume one needs to produce the video but doesn't really understand too much about codec options
• Multiple targets per video: add multiple output formats/ devices/ resolutions to each 'master', which implies:
• Ability to scale and crop per-target
• Work with multiple input streams, including image sequences, audio and subtitle tracks
• Constraints not just presets (per target) This one is a bit weirder, but what if you didn't just lock down all options, but limited them depending on the target device - for each target, expose only those knobs that would vary quality but not create an unplayable file. Optionally, one could add filesize/ bandwidth/ CPU horsepower to the constraints as well as the target device/OS
• Live preview: encoding is usually happening at a reasonable frame rate, anywhere from 1 to 20 frames per second. It should be possible to display (selectable with a timeline scrubber) a frame of the video and live change the result of tweaking the available knobs (in the ranges specified by the constraints
• Easily save settings for an encode as a new batch for future videos