The present and future of post production business and technology

Assisted Editing – the beginning

I originally wrote this to a friend over the weekend and I thought that, with a bit of an edit, it might be interesting to some others.

Last week we released two new pieces of software through my long-time company, Intelligent Assistance, Inc: First Cuts and Finisher. Collectively we call the category “Assisted Editing” because they, well, assist the editor and going with Assistant Editor was getting confusing.

What I think makes this so exciting is that it’s the first real innovation in editing since Non-Linear Editing was popularized with the release of Avid’s Media Composer version 1, 19 years ago. 19 years is a long time in the computer business. Non-Linear Editing has certainly developed since that first release of Media Composer. Then 160 x 120 16 gray images were the “state of the art” while now we effortlessly manipulate High Definition (and beyond) images in our systems. But, despite this development, the way NLE systems work today is much the same as they did 20 years ago.

What Media Composer did, was to make the “cost” (time, effort, expense) of an edit much more affordable. With linear editing bays, making small changes to, or alternate versions, of a program was difficult (read expensive). Non-linear dropped the cost of a change, revision or alternate edit dramatically. Assisted Editing achieves similar dramatic cost reductions by allowing the computer (and specifically our software) to do some of the work of the editor.

Back at NAB in April we announced “The Assistant Editor” for long form documentary editing. After three months of beta testing we have now released that application as “First Cuts”. First Cuts (for Final Cut Pro right now but we’re exploring other platforms) take the log notes made by a documentary editor (or their assistant) and turns them into very, very fast first cuts. This allows editors and producers to explore the stories that are available in the material, and to juxtapose different versions while they seek inspiration and direction.

First Cuts take the log notes that long-form documentary editors have traditionally made (and their workflow) and makes them much more useful. The logged clips are exported from FCP as XML and opened in First Cuts, where the editor chooses opening title and lower third template (Motion templates preferred), the duration and story keywords that will be used for this edit. Within seconds it creates an edit with beginning, middle and end to the story arc, with opening title placeholder and fully finished Lower Third titles used appropriately to identify speakers on camera. The edit will also have b-roll used appropriately and the audio on the b-roll is lowered in volume and faded in and out. It’s an edit a producer or finance person can watch without being distracted by jump cuts or lack of visual interest from the absent b-roll. You can see the process in a quick five minute tour at the First Cuts home page.

First Cuts is primarily focused on long form documentary editors because they have the discipline to enter log notes (if they don’t they will have a very hard time of it) and because they need to explore a lot of material. The payback is significant. Fortunately there are a lot of documentary editors working with Final Cut Pro.

We discovered that one way of working with First Cuts would be to find a cut that was close to what was desired, and then the skilled editor would add the polish, trim and emotion that a first cut lacks. In doing that we discovered that blowing away the b-roll and titles was the cleanest workflow. Since that removes a lot of the time saving, we decided to make a product that restores that finishing effort. Hence, Finisher.

Finisher takes a project with an edited a-roll sequence (aka “radio edit”) and adds Lower Third titles and b-roll as above. This is the perfect bookend product to First Cuts.

But we worked out that we could do a lot more with Finisher for those who don’t want/need to enter a lot of log notes. Finisher will work with any of the log notes that are provided for First Cuts, but does not require them. In fact Finisher can randomly choose b-roll to place in a Sequence. Location of b-roll can be forced with Sequence Markers, so obvious jump cuts are covered first. While it can run with random selection it will use b-roll search terms in the comment field of those Sequence Markers and search for matching b-roll if available.

You can see some examples of Finisher’s work, including a guided tour that shows a couple of different ways it can work, at the Finisher home page. Particularly interesting is the final example – ‘Ode to the Beach’ – because that is a simple audio recording, some b-roll pulled out of my stock collection from Final Cut Server, and Finisher adds b-roll to the cut (with Sequence Markers) randomly. The result – for almost zero effort – is quite acceptable and may even be useful in some situations.

Not too shabby for a version 1 product.







One response to “Assisted Editing – the beginning”

  1. Thanks for this detailed post, provides a clear picture. I agree that ‘the Assistant Editor’ is an important application, and one of the early true innovations in ‘semantic video’. Our product Frameline 47 [] has some similarities in terms of content taking the lead, but with a different approach and the additional objective of creating an ongoing video ‘asset’ from the logging process.

    With Frameline 47, we designed an annotation model based on MPEG7, with files broken into segments and groups which equate to shots and scenes. Each of these units can then be tagged (at a basic level) with Character, Plot, Location, Time & a value. With this schema, you could then automatically re-purpose video content into strings, for example, with a Soap, you could generate a sequence consisting of a series of condensed plot summaries, or follow a particular character, or set of characters interaction over a period of time.

    In a sense then, this evolution might end up giving control to the viewer, where video is actually navigated according to its content, rather than its timeline.