Issue link: https://digital.copcomm.com/i/1546054
M OT I O N P I CTU R E S O U N D E D I TO R S 119 moving from 5k to 12k. I also did experiments on how the ear uses nonlinear processing to use combined frequencies such as 2.2k, and 2.4k, all with wavelengths shorter than the head, to generate time delays between the ear that behave like a 200 Hz sound and cause the perception of an offset from a centre position in the horizontal sound field. After four years of research, I moved to London, wrote my dissertation, and eventually got my PhD, but I needed to find something to do… CM: So with that said, how did you transition into doing work in the pro audio community? JB: This question reveals how visualising a potentially profitable future in which you're actually too early to be one of the first people who could replace existing established technology with new emerging technology and solutions that haven't yet been proven appropriate or suitable is not a good business start. In 1978, after my stint in Signal Processing at the PCL, a friend who had recorded the music Pam and I did in Cleveland was now running a mastering company (with a Neumann cutting lathe) near Chicago who offered to hire me to explore the idea I had of creating an all-digital recording facility. As it turned out, I knew that Bell Labs had developed some DSP boards that could plug into the somewhat improved minicomputers of the day, and this was the core that made this concept seem possible. As things developed—I got flipped off this direction (fired) but learned in the experience, that 1) it turned out they were right to do that at that time because it would have been unfeasibly expensive, and I also got bounced into attending a DSP conference where the head of a new Digital Signal Processing Group (DSPG) at the Polytechnic of Central London (a 4-mile bike ride from where I was living in London) invited me to pop in and visit the group. He wound up getting me a job there, and I was granted the opportunity to not only learn about signal processing, but also had access to reasonable A to D and D to A converters to get audio in and out of the computer. I spent nearly 10 years there, and over that time had various jobs and grants in which I explored and published academic papers on. CM: Jeff, could you describe the conceptual origins of WordFit, and how you came to see the power of the WordFit concept in the pro audio world as a creative tool? JB: Once I was working in the Signal Processing Group (SPG) in the late 1970s, I started to use a 32-bit minicomputer to explore hearing- related problems such as "can humans hear the phase changes introduced by anti-aliasing filters?" or "can a de- reverberation algorithm make it easier to hear speech more clearly?" I also did a project for the BBC to automatically take rapid and frequent loud microphone pops out of a live concert recording with a famous violinist, so it could be broadcast without those offending sonic distractions. As well, since I was still angry about getting fired from the digital studio project, I wrote a long article about what I had learned about the possibilities of digital audio studios,* (which for those interested, can find here: Bloom, P. Jeffrey. (1979), Studio Sound, "Into the Digital Studio Domain," Part I April, pages 56-60 and Part 2 May, pages 58-70.) The thing to certainly recognise is that the article was a snapshot of what a "Digital Studio" might look like before Digital Sound Studios existed at all… After that article was published, I received a call from a reader, Garth Marshall, who worked as a sound recordist for films, including notably, Monty Python ones, and we spoke about what was going on in the film sound industry. He also happened to live near me. (I even wrote a program to create "beast-like creature voices" for Terry Gilliam, who came to listen to them.) Two years passed before Garth mentioned to me the process called ADR—"Automatic" Dialogue Replacement (now known as Automated Dialogue Replacement), which is, by no means "automatic" or "automated." It took a lot of actor skill and extremely detailed and time- consuming tape editing to achieve convincing lip-sync. That is when, having had just enough experience with Digital Signal Processing, a knowledge of hearing, and an understanding of speech signals, is when I wondered if this was a problem that could be achieved in a computer. I also knew the alignment algorithm itself was not going to be trivial to create. In this pursuit, I went to a speech processing conference in London, and heard a talk by John Bridle on improving speech synthesis by looking at the timing of speech features of synthesised speech relative to the same features in real speech. Bridle used a very efficient feature alignment algorithm which we were able to adopt to our purpose and needs. With that, I broke the processes "…it might be amusing for some to know that the name Titan was actually a little 'joke' based on the Fix-Sync function (that no one ever got) because we were "tightening" the conformed audio."