Is beat detective dead?
Once hailed as game changing and certainly one of Pro Tools’ most buzz worthy features, has beat detective officially reached legacy status?
Yes and no. While elastic audio and tick based time references for audio tracks have made life easier for many editors, in some instances literally cutting the time it takes to complete a task by ten fold or more, I firmly believe beat detective still has a place and time (no pun intended) in a Pro Tools, at least for now.
BD or Elastic Audio?
Essentially, all beat detective really does is “detect beats” or transients of audio material (usually drums, percussion or monophonic bass/guitar). Once these “triggers” have been detected, the user has some options. Traditionally, slicing or separating a larger region’s transients into individual smaller regions and then subsequently “conforming” or quantizing these slices to the grid was beat detective most commonly used workflow.
This usually works OK, but can manifest problems worse than the bad timing was in the first place. Because beat detective physically cuts and moves regions (thus maintaining absolute phase accuracy in multi-track scenarios like drums) the gaps resulting after a conform are a neccissary evil of the process. While these gaps can be filled and cross-faded inside beat detective, many times the process of trimming the conformed regions reveals the decay of the previous note (this is especially the true when notes are extremely rushed or ahead of the beat, or when you are trying to add swing to an originally straight feeling performance). Often what engineers are forced to do is manually address these problem spots one at a time, usually pasting an alternate hit over the un-trimmable section (either from another part of the session or from samples taken from the kit during recording). This can take several hours, depending on how out of time the performance is and how often manual intervention is required. I personally have had sessions where completely rebuilding a section of drums manually with samples from other parts of the session was a faster option than trying to use beat detective. At which point I usually say to myself, “It would have been cheaper to rebook the studio, hire a better drummer and re-record the drums.”
Much of what was once done in beat detective has been replaced by the elastic audio feature set. Because elastic audio “warps” or stretches notes into time, no gaps are created when quantizing or conforming audio to a different feel, and since the “rhythmic” elastic audio plug-in is essentially just adding or subtracting space in between hits, you generally don’t have to worry about the “zipper” artifacts that plague granular re-synthesis based warping.
Some engineers argue that unlike beat detective, elastic audio can’t maintain perfect phase coherency across multiple tracks because the algorithm processes each track uniquely. While this is true, you can get around this somewhat by grouping the tracks before elastic analysis and processing, this method will try to maintain phase coherency as much as possible as audio is stretched. I generally find this group method yields usable results almost every time, and while I do hear very subtle differences in the phase alignment of the kit, they are so subtle that the cost of doing a 4 to 8 hour beat detective session wouldn’t warrant the improvement (nor be affordable or in the best interest of the client).
Like beat detective, I find that working with small sections (4-8 bars) at a time yields the most accurate results when conforming with elastic audio. This allows me to focus and address any errors manually as I work through the song. I generally do not try to conform fills and opt to manipulate them by hand, as the timing during these sections is often dramatically different than a grid quantize would suggest. For bass, I generally avoid automated conforming altogether and adjust each note by hand with warp markers.
All that said, I still haven’t fully addressed the question, “is beat detective dead?” For quantizing drums and other percussion, pretty much. At least for my workflow it is. Simply put, time is money and elastic audio saves me lots of time and sounds really good if you know how to use it. But I still use beat detective almost daily for it’s lesser known feature, “groove template extraction.” Groove template extraction uses the same process of detecting beat triggers but instead of separating/conforming those triggers it records them to a groove template. A groove template is essentially a list of tick offsets from the grid, with a goal being to extract the groove of one things and apply it to another. For example, if the “and of four” is slightly behind the beat, a groove template would record the exact number of ticks that note was from the grid point (…|4|480). Subsequently, when I quantized something with that groove template, the off-beat of 4 would be pushed back a bit. If I had to imagine a groove template as a text file, the following 1 bar template would describe a swung 8th note feel:
Grid: 1|1|000 Offset: +0 ticks
Grid: 1|1|480 Offset: +100 ticks
Grid: 1|2|000 Offset: +0 ticks
Grid: 1|2|480 Offset: +100 ticks
Grid: 1|3|000 Offset: +0 ticks
Grid: 1|3|480 Offset: +100 ticks
Grid: 1|4|000 Offset: +0 ticks
Grid: 1|4|480 Offset: +100 ticks
When beat detective is in groove extraction mode, it measures the difference from the triggers it detects to the tick grid and records them either to the groove clipboard or a new groove template. At that point, I personally would use elastic audio to quantize another region with the newly created groove template.
In this workflow, there is a sort of symbiotic relationship between beat detective and elastic audio, at least until Avid integrates groove template extraction into the elastic analysis engine (I believe sonar has a feature like this in its elastic implementation). I imagine an elastic audio system that would allow me to let’s say, click on a menu and see a list of every other elastized track and choose “quantize to follow track XYZ.” At which point the warp markers of one track would reconform to match the analysis (or warp markers) of another track in the session.
Bar Beat Marker Generation
One additional usage of beat detective is it’s bar|beat marker generation mode, which is specifically designed to conform a session’s grid to a freely recorded piece of audio. It is sort of like reverse groove template extraction, where audio that may not conform to any one tempo is assigned a grid based on the transients absolute sample position. Think of a performance and it’s transients representing specific beats of the music (1,2,3,4 etc). Now imagine the grid is flexible and all you have to do is place a push pin at each transient marking what bar|beat it represents. In this scenario, Pro Tools switches the tempo ruler to bar|beat marker mode and every pair of bar|beat markers may represent a change in tempo. Remember, you are not conforming or quantizing the audio, it is the grid itself that is moving. A freely recorded piece of music that may sound consistent will still have small tempo changes, maybe a fraction of a BPM per measure.
Most would ask, “why on earth would I want to conform the pro tools grid to my audio, and not conform my audio to the grid?” Say you have a session that was not recorded to a click, but you like the feel and want to add additional MIDI and/or do some editing. Your options are either to work completely in slip mode (without the ability to quantize additional MIDI performances), or to develop a bar|beat grid based on your freely recorded audio. The people who track notes for guitar hero and rock band have to do this kind of stuff all the time, as most of those classic songs from the 70s and 80s were not recorded to a click and creating an accurate control track requires MIDI with many tempo changes over the course of the tune.
An alternative to Beat Detective’s bar|beat marker generation is Event>Identify beat, which uses the same concept of sample based bar|beat markers to develop a grid around a freely recorded session. The only difference here is that you will use tools like tab to transient to place the cursor and identify beat to manually insert a bar|beat marker into the session. I actually prefer this method over beat detective as it gives me a lot more control and I can generally get a song mapped (placing one marker per measure) in under 10 minutes.
While the sonic purist may still consider using beat detective for mult-track drum timing, and certainly every unique situation calls for a slightly different approach, over the past 2 years elastic audio has definately taken a big bite out of my beat detective work flow. I will of course continue using beat detective for groove template extraction until a faster solution is implemented, and I will probably always use indentify beat when mapping out free performances, but for the most part I am glad to leave those aweful gaps and tediously long editing session behind.