Scratch - A Robust DI Solution On An Indie Budget

Last night I went to the very swanky Groucho Club in Soho. I would normally not be cool enough to get in, but last night I had an invitation. After clearing the barricades at reception, I sauntered past the smattering of celebs and headed upstairs to the Mary Lou Room.

I had been very invited to checkout Assismilate's DI conform and grading solution, Scratch. I had heard a fair bit about Scratch on various sites including HDforIndies and there is a good case study at Studio Daly as well. But I had yet to see it for myself in action.

Like many, I have been more and more involved with FCP as an NLE over the past few years, so I naturally focused my attention on what looked to be the wholly grail of affordable DI and NLG (non-linear grading) solutions. Running on a mac, Final Touch, provided a XML based, tapeless workflow with a UI that wowed many experienced colourists. Having assumed the price you pay for Final Touch is a direct reflection of an intention to work with a company as they develop a product that has tremendous potential, I have regarded the acquisition of Silicon Color (the product's manufacturer) by Apple with trepidation. Will there be further development of the product that many companies have invested time and money in, or will Apple use the R&D to further their product line to match their wider pro-sumer strategy? Ultimately, if no further development is done on Final Touch, the customer's who purchased the product will have been royally shafted. The product has great features, but lets the user down greatly in the workflow it promised.

It is from this skeptic place, that I ventured out to look at a similar offering in Scratch. What I learnt though, pleasantly surprised me. Assimilate have not tried to promise more than what is feasible and realistic. Scratch is a cost effective, data centric, grading solution much like Final Touch, however, the product is focused and asks the user to work its way to achieve the best results. It also goes beyond grading and offers a balanced solution to meet the needs of Data Conforming for film based productions.

Scratch really is a film tool. It works predominantly with DPX files and has been optimized to use them well. It now offers Quicktime support. Here, Mike at HDforIndies breaks that down, along with issues involved like the need for comprehensive file support and workarounds for unsupported codecs. During my demo, there was a QT rendered out from the offline with BITC that was used to frame match the conform. The system seemed quite happy with both the QT and the DPX.

Scratch as a product is modular and sold as a multi tiered license. The facets include:
1) Project
Data management
2) Process
Image processing and rendering including plugins
3) Edit
Conform and simple editing tools
4)Extensions
Additional Tools
5) Media Layer
Quicktime export etc.
6) Color
Grading
7) Scaffolds
Secondaries

For me, seven modules is overkill for this product, but I am sure there is reasoning behind it. One nice thing is that because it is modular you buy separate licenses and do conforms in a back room, while saving your plush client suite for grades.

Scratch is file based, optimized for DPX (handles 2k DPX in real time) and resolution independent (HD, 2k, 4K).
Includes a DPX player as a stand alone app which lets you preview scans.

Project
With DPX files, scans are individual frames or images, like a tiff sequence. In Scratch, groups of frames are called 'shots'. Demo loaded a good size group of shots into project from directory very fast and allowed navigation, playback and work on shots seemlessly as you would expect from an NLE. But these are 2K. As Scratch can handle various sources, to avoid processing issues you can, for example, 'soft load' jpegs and converts it to DPX. The way is manages the robustness of power needed is by offloading much of the processing work onto the graphics card. I think a lot of products are doing more and more of this because the power is there to be taken advantage of. As a result, a better gfx card equals better performance.

I believe scans are done in a way resulting in metadata associating takes of shots, which are grouped by staking them in the project viewer.

In order to conform, the op would create a "Construct" and an "EDL Load". There is some XML functionality in the product, but my demo didn't cover this at all. I think customers are building there own XML based workflows by writing their own code, much like with Shake. The workflow utilises the tried and tested route of applying EDLs to the shots and woop, your done! 55 shots were conformed in less than one second. You can then import your BITC (burnt in time code) quicktime and frame match the newly conformed 2k material to your offline quicktime. These two nicely stack in the timeline and we can generate timecode and clip details on screen against our conform sequence as well. We can also view in split screen and slip shots as needed to fix discrepancies in the EDL. The media is loaded rather than imported which apparently makes a difference in how fast it runs, but to me this seems like semantics and just take it as read that it works fast. And finally, there is sound!!! unlike Final Touch, Scratch supports sound sync. Not as you might hope however, they're working on it. At the moment you can render out an AIFF or WAV and slip it into position on the timeline. Soon to come, BWAV with metadata and quicktime Audio Support. In my demo, the sound didn't actually work, but I will take their word for it.

Grading
Within the primary grading room there is a side bar for stackable versions. For example, there could be multiple versions of the grade, multiple takes from the the shot that need to be looked at, the rendered QT, or variations of VFX. This is very handy.
Along the bottom of the grading screen lives a tray for the shots which quickly allows you to apply grades across a selection or indeed all of the shots.
There are lots of other groovy features, but I think they are pretty much in line with the rest of the markets' offerings.


Scaffolds
Secondaries can be stacked so they apply on top of other vignettes or can be set to cut through other grades to the source image. This seemed like a sort of compositing feature in the grading. Tracking is similar to Final Touch, but the track button was missing from the demo system. Key point: know what you want to do and buy a license that allows you to do it.
As for performance we had 3 secondaries and a primary on 2k and had real time playback. What Scratch does cleverly is use the GPU to process on the fly as grades are applied, they are in effect rendered. When finished all that needs to happen is write the final files as the processing has been done. Simple test I saw laid a single primary across 2268 frames and was processed in under a min. 2268 frames = 1.5 min. So, basic processing will happen faster than real time. This is impressive, but I wouldn't be sold until I saw a real life project put the system through its paces.


Extensions
Can use full range of Foundry plug-ins, Ultimatte, etc. I didn't see this in action, but I gather most of this stuff is better left to other applications like Shake or similar.

Hardware
Hardware wasn’t really covered in the session, so I have copied this from the Assimilate site.

The performance requirements listed below are based on a file-size of approximately 12MB per frame, which equals a 10bit log DPX file at 2048x1556 pixels resolution (referred to from now on in as 2k):

a) For 25fps playback of 2k 10bit DPX files the technical minimums are:

25 guaranteed I/O transactions per second for a total request size of 12MB per transaction

Minimum latency per I/O better than 40ms

Sustained read bandwidth across array, minimum of 301MB/s

This is the absolute minimum that MUST BE sustained at all times.

b) For dual stream 24fps playback of 2k 10bit DPX files:

48 guaranteed I/O transactions per second for a total request size of 8.4 to 12MB per transaction

Minimum latency per I/O better than 20ms

Sustained bandwidth across array for dual stream operation at 24fps per stream, minimum of 577MB/s

The performance of suitable direct-attached arrays for SCRATCH will usually lie somewhere in between these two boundaries [of a) and b)], whereas SANs may often exceed this performance envelope. The smaller request size of 8.4MB is roughly the size of a Super2k DPX frame (2048x1080 at 10 bit log). While Super2k will be easier on the bandwidth requirements; it will still require full I/O transaction performance. Sustaining these I/O transactions tends to be more of a problem for SATA arrays at this point.

Tower recommended specs:
BOXX 8212 H I/O station (dual Xeon)
BOXX 8312 RTX I/O station (dual-dual core Xeon)
BOXX 7512 RTX I/O station (dual-dual core Opteron)
BOXX 7412 I/O 3D station (Dual Core/Dual Opteron)
BOXX APEXX HS I/O station (quad-dual core Opteron) (Coming soon)

Globalstor ExtemeStor DI Station

HP xw8200 Workstation* (dual Xeon)
HP xw9300 Workstation* (dual Opteron)

IBM IntelliStation A Pro* (dual Opteron)

* Requires external storage

Color Management
Assimilate Scratch currently supports 3D look up tables (3D LUTS) for a range of color management solutions, ensuring a calibrated and truthful color manipulation throughout the DI chain.

Arri CMS Color Management System
Imagica GALETTE Color Management System
Kodak Display Manager System
Pandora Pogle Color Correction Systems (Supported LUT formats a3d, m3d)
Rising Sun Research cineSpace Colour Management

Bottom Line
Software packages run from £10 - £20 K. I should have more details emailed through to me and will update soon.

They have 5 systems working in London and 'loads' in the States.

Final Thoughts
Good stuff in all. I really liked the interface, it straight away reminded me of a Discreet product (Flame or Smoke). Really slick, but usable. I felt that with a bit of poking around, I could get to grips with the functionality of it. This is important as it would need to be used by Technical Assistants in order to conform, but also needs to meet the demands of highly skilled Colorists. Most functions ran smoothly on the demo system and in all I could see it fitting nicely in a number of setups.

The proof is in the pudding. Mike Curtis recently said about new tech solutions "the sales pitch sounds the same whether it works great or is a clusterf*cked, non-functional cripple of an app." I think there are those of us who are a bit more skeptical having suffered with technology that hasn't met expectation. Final Touch was/is a great app, I work with Pogle Colorists who are really eager to jump on the Final Touch. But it's detractions were not apparent at any demo I attended. The problems are all in the workflow.

I get the feeling those who have moved from Final Touch to Scratch have been happy with the decision. I also think Scratch is not built on a promise. It does what several other products do (Nucoda, Baselight, Final Touch, Lustre) . Like those other products, Scratch has it's strengths and weaknesses. I was told Baselight is a better grading tool, not surprising when you look at the Blackboard it comes with. Scratch has tried to be more evenhanded devoting much effort to the strength of its conforming tools similar to the way the Nucoda system does.

Baselight Blackboard

Ultimately, this is a film product. The broadcast market and HD markets are getting largely ignored by products at the moment. Perhaps it is down to confusion about the fact that in video work there is no need for a digital intermediate. There is no intermediate stage in video post. There is however a real desire for the types of tools and potential workflows that non-linear grading can offer. Scratch I don't think is really answering this call. They are staying true to form, rather than repackaging to meet extended market share. So, if you are shooting a good bit on film, or mixing film and video, then this could really be a useful addition. If you are a facility looking to develop the DI side of things this may be a cost effective solution. But traditional HD conforms are probably better done in a FCP, Smoke or Avid. That still leaves the colour side of the system, within which there are really strong tools that most Colorists would be happy to work with. The system is fast and robust and with further video file integration it could be a strong offering for video post houses as well as those moving into film /DI post.

Finally, thanks to everyone at Assimilate for taking the time to show me the product.