Pitch Changing

Aligning Dialog

When working with dialog, it might be necessary to adjust the timing of certain lines to compensate for camera moves, tricky edits, or even ADR recording.

When camera angles change in a scene, sometimes directors will use a person's line from another camera angle, and therefore a line from another take. This alternate take might overlap images of the character performing the same lines from the original take. When this happens, editing between the two takes right when the angle changes might work, but often the continuity is lost and the first take must overlap the image of the second take. If luck is with you, the lip-sync will match. If not, you might have to alter the take to match the image. Likewise, sometimes an actor is not able to recreate a performance in the ADR studio that matches well to the picture. In a case like this, you might need to alter the dialog audio to match the original.

Plug-in tools such as Synchro Art's VocALign plug-in are created for matching the timing of one recording to another. This plug-in can analyze an existing vocal take and conform a similar one to match the timing of the original. This can help a problematic ADR session by taking a good performance and aligning it with the original take.

As you can see in Figure 7.17, the VocALign plug-in has two windows in it. The top window is used to analyze the original or reference audio, and the lower window is used for the audio that will be processed to match. To align one dialog take to another, follow these steps:

1. Choose Audiosuite > VocALign to open the plug-in. This is an Audiosuite-only plug-in and will not be found in the mixer's inserts .

2. Select a region that contains the dialog take you want to reference. This region will be used as a guide to which the new take will be aligned.

3. Press the top guide button, indicating that the file to be analyzed is a guide track.

4. Press the Guide button. A waveform display of the selected region should appear in the Guide window, as seen in Figure 7.17.

5. Select the region that contains the new dialog that you wish to align.

6. Click the Dub button to indicate that this is the audio to be processed.

7. Click the Align button. An outline of the processed file will appear above the guide track indicating where each peak and valley will line up in the new file, as seen in Figure 7.18.

   Figure 7.18. Once the Align button has been pressed, this outline appears above the guide track showing where the new file will line up.

   

   
   

8. In order to hear the result, press the Preview button.

9. Click on the buttons in the flexibility setting along the bottom of the window to see which one yields the most accurate match to the guide. You will have to press the Align button again after choosing another flexibility setting.

10. Decide whether or not you want the new file to replace the unprocessed file. There are several choices. You can leave the processed file in the Audio Regions list without placing it in a track, you can place the new file on another track specified in the Destination pull-down menu, or you can replace the file on the same track. Create a duplicate playlist and append the name with "-align" to indicate that you have aligned all the regions that were necessary and that the track is complete. This way, you can quickly return to the unmodified version of the track if you need to.

11. Click the process button. Done.

As with any plug-in, overuse is not a good idea. It is always best to get a better performance than to "fix it in the mix." VocALign is a tool, not a crutch.

Noise Reduction and Restoration

While editing dialog, you might find that the recordings have unwanted noise on them. In lieu of re-recording them in an ADR session, it might be possible to clean up the tracks using plug-ins that analyze noise content and filter it out, leaving the dialog relatively untouched. Digidesign's DINR and the Waves Restoration Bundle are two popular systems for removing noise. Sonic Solutions NoNoise is one of the oldest developers of this type of technology, and it, too, is now becoming available to TDM systems. Sonic Solutions and Waves plug-ins have more sophisticated tools for removing hum and clicks, but all three have a similar broadband noise reduction plug-in that will analyze the noise content of a signal and create an inverse noise-canceling algorithm to remove the noise while leaving useful program material intact.

You must have a section of audio without any dialog that contains the same type and level of noise as the areas you wish to clean up. A second or so of background noise will usually do nicely . You should be able to find a section of this background noise prior to, or just after, a take. You might have to return to the source tapes to find such a section. Most professional production sound mixers will record long sections of background ambience for use later on in dialog editing. While these recordings are primarily used to create seamless background ambience for sections with many dialog edits, they can also be used for noise analysis and removal. Here's how the process works using Wave's X-Noise plug-in. The process is similar with Digidesign's DINR and Sonic Solutions' NoNOISE.

In this example, you have some production dialog audio that was recorded on a set using several fans to create wind for the scene. These fans made a decent amount of noise, which was recorded through the boom mic. The director would like to keep the original dialog recording for this scene because the performance was very special. You must remove as much of the fan noise as possible in order to make this recording usable. Here's how:

1. Instantiate the Waves X-Noise plug-in on the dialog track in question.

2. Find and select a section of audio that is just the background noise without any dialog. A second or two will do.

   NOTE

   NOISE SAMPLES

   The noise you select is critical. Analysis of the noise sample will be used to cancel noise in audio that has dialog in it. If the noise floor changes between shots in a scene or even between different takes in one shot, reanalysis will be needed in order to accurately remove the noise. Changes in the noise floor could be from movement of the fans, in this example. Air conditioners will often create noise that changes when the unit cycles on and off. Make sure the noise you analyze is the same noise you want to remove.

3. Open the plug-in window and press the Learn button at the bottom. The button should flash yellow and say Learning, as shown in Figure 7.19. The plug-in needs to analyze this noise sample in order to create a profile that will be used to remove the noise. This is what the Learning mode is for.

   Figure 7.19. Waves X-Noise plug-in set up in Learning mode in order to analyze the noise sample.

   

   
   

4. Make sure that both pre- and post-roll are disabled. This ensures that when you play the selection, only the selected audio will be analyzed by X-Noise. Any audio outside your selection might contain desired program material and should not be part of the noise profile.

5. Play the portion of audio. You will notice a spectral display of the audio inside the X-Noise window while playing.

6. After playing through the whole selection, stop playback and press the flashing Learning button again to end the analysis. In the spectral display you will see a noise profile that represents the noise components in that portion of audio.

7. Play the section from which you wish to remove noise. Putting it in Loop Playback mode (Command+Shift+L) might help while adjusting the X-Noise parameters. If you have other tracks playing along, you might notice that the one with X-Noise on it is severely out of sync with the rest. This plug-in has a large amount of latency. It takes more time to process the audio, and therefore it is delayed in playback. For now, just solo this track so as not to be distracted by its lack of sync.

8. Adjust the Threshold (Thresh) slider up. As the threshold increases, more noise will be removed. The function is similar to a noise gate or downward expander. The threshold is the level at which the signal will be allowed to pass through and be heard . If the signal is below the threshold, it will be gain-reduced. The amount of noise reduction is determined by the Amount slider. Adjust these two parameters until you achieve the desired noise reduction. As noise reduction increases , you will see two moving lines and one stationary white line in the spectral display. The white line is the noise profile curve adjusted to the threshold setting. When you raise the threshold, the noise profile curve moves up. The red moving line indicates the signal input to the plug-in. The green moving line is the result of the plug-in's processing. In sections where there is only noise, the red line should appear below the green, indicating reduction of the noise floor, as seen in Figure 7.20.

   Figure 7.20. The X-Noise plug-in in action. The green line shows the result of processing; the red line shows the noise floor being reduced from its original level.

   

   
   

   NOTE

   HOW X-NOISE AND OTHER NOISE REMOVERS WORK

   X-Noise uses a large number of downward expanders on individual frequency bands to achieve this quality of noise reduction. The noise profile sets relative thresholds for each frequency band . The Resolution setting determines how many individual bands there are. The higher the number of bands, the more precise and accurate the noise reduction will be. The trade-off is that the higher resolution creates larger amounts of latency. The method in this example will avoid the latency issue by processing the files offline with the Audiosuite X-Noise plug-in. With that in mind, using High resolution will yield better results in most cases.

9. The High Shelf frequency and gain parameters allow you to adjust the threshold of higher frequencies in the noise profile. The higher frequencies in dialog can sometimes be close to the same energy level as other noise components in the signal. Reducing the threshold at these frequencies by entering negative values for the gain parameter can help maintain clarity when performing a large amount of noise reduction. Raising the high-frequency threshold can help reduce noise components in the more audible ranges when the program material itself is mainly in lower frequencies. The result of this equalization can be seen as changes in the shape of the white line. Figure 7.21 shows the same noise profile as in Figure 7.20 with the addition of -30db of gain with a corner frequency of 4006Hz. Notice the right side of the white line is lower. The threshold of frequencies above 4006Hz has been reduced by a maximum of 30db thus allowing more high frequency detail to come through.

   Figure 7.21. The noise profile has been adjusted by the High Shelf parameters. Notice the right-hand side of the noise profile is lower than in Figure 7.20.

   

   
   

10. The Dynamics section of this plug-in controls the speed at which the noise reduction works. The Attack parameter determines how long in milliseconds it takes for noise reduction to occur once the signal goes below the threshold. Conversely, once the threshold is exceeded, the Release parameter determines how long it takes for noise reduction to cease . Quicker attack and release times will improve clarity in most situations. However, the quicker these times are set, the more noticeable the processing becomes. At very fast settings, the processing starts to sound rather robotic and mechanical.

    NOTE

    LISTENING TO THE "DIFFERENCE"

    The Audio and Difference buttons located at the bottom right of the plug-in window can be used to hear exactly what noise is being removed. When you click the Difference button, you will hear only the noise components that are being removed. Anything you are hearing will not be in the processed signal. This is a good way to check and see if you are removing useful program material. If you are listening to the difference and hear any clear dialog, your settings are too severe. Try backing off the reduction and lowering the threshold.

11. Once you have adjusted all the parameters and come up with a setting that achieves the best noise reduction versus sound quality, you must save the setting as a preset. This preset will be used to perform offline processing using the Audiosuite version of X-Noise. This offline process eliminates the latency issues that arise when using X-Noise in real time as a TDM or RTAS plug-in. To save the preset, choose Save To New File from the Save submenu in the upper-right-hand corner of the plug-in window, as shown in Figure 7.22. It is also possible to save settings from the standard preset menu in the upper left of the window, but with Waves X-Noise, the only way to save the parameters and the noise profile is to use this other submenu. Try to use a descriptive name for the preset that you can identify as applying to this particular section of audio. You might be saving many more noise presets before an entire film is completed. Being able to refer back to these settings can be very important if you decide to alter the settings at a later time.

    Figure 7.22. Choosing the Save To New File option under the Save submenu.

    

    
    

    NOTE

    USING X-NOISE IN RTAS OR TDM

    If you want to use the RTAS or TDM version of X-Noise, you will have to compensate for the extreme latency that this plug-in will create. Please refer to the plug-in latency compensation section in Chapter 8, "Mixing to Picture in Pro Tools."

12. With the preset saved, you are now ready to process the files offline. Choose Audiosuite > X-Noise to open the Audiosuite version of X-Noise.

13. Click the Load button and choose Open Preset File. Locate the preset you saved in Step 11.

14. Click the Load button again, and you should see the name of your preset file listed in the middle section of the submenu, as shown in Figure 7.23. Select the Both option to load both the noise profile and the parameter settings from the preset file. Your plug-in should look the same as it did in the RTAS or TDM version.

    Figure 7.23. Once a preset has been loaded into the Audiosuite version of X-Noise, you have three choices available. You can choose to load the noise profile by itself, the preset by itself, or load both at once.

    

    
    

15. Before you process these files, you should create a duplicate playlist in which the processing will occur. This gives you a quick way to return to the unprocessed files in case you wish to make a change. Understand that once you process the files with any Audiosuite plug-in, only those portions of audio that are visible inside the regions will be processed. The parent files these regions belong to will not be entirely processed. It will not be possible to "open up" the regions with the Trimmer tool once they are processed. Using an alternate playlist allows you to go back to the originals , which will retain any handles they might have had, allowing you to tweak edits if need be.

There are other plug-ins that help remove different types of noise. The Waves Restoration Bundle has X-Hum, shown in Figure 7.24, which is designed to remove unwanted hum generated by electrical grounding problems, portable generator noise, and even air conditioning and fan noise. Most steady-state noises like these can be dealt with using X-Hum. Using a series of tight notch filters placed in the harmonic series, X-Hum attacks not only the primary tone of hum but also all the subsequent harmonics it generates. X-Hum's controls are very intuitive and can be used in the same way as in the X-Noise example above. Typically, removing hum would be done prior to broadband noise reduction. When performing offline processing, simply use the X-Hum Audiosuite plug-in first, and then apply X-Noise to the files already processed by X-Hum.

The X-Click and X-Crackle plug-ins are primarily designed for vinyl record and optical film sound restoration and are of limited use in video post-production.

Noise reduction tools like the ones presented in this chapter are powerful and complex. Experiment with some noisy material to get a feel for how each one works before processing critical files. It is often a trade-off between the amount of reduction realized and the amount of artifacts introduced into the de-noised signal. In some instances, you will find that one plug-in works better than another on a particular piece of audio. In other instances you may find that processing the signal twice with slightly lower amounts of noise reduction in each pass will result in a greater total noise reduction with fewer artifacts than you could get with a single pass. The controls are very intuitive and can easily be mastered with experience. The ability to remove much of the unwanted noise can drastically improve the resulting soundtrack, leaving more space and dynamics while improving clarity.

Surgical EQ

Sometimes the dialog tracks will have small anomalies in them, such as mic "thumps," where someone accidentally touched the microphone or the boom operator was unsteady and bumped the microphone against something else. If this occurs during a word or phrase, it could potentially make that dialog recording unusable. It is possible, using Audiosuite plug-ins, to process just a very small portion of the audio file in the affected area to treat these anomalies.

Using the example of the mic thump, you can select the affected area of the sound file and then open an Audiosuite EQ plug-in that has a high-pass filter in it such as the Waves Q1. Using the preview function, adjust the parameters of the Q1 to filter out the low-frequency thump in the sound. You will not be able to entirely eliminate the thump sound, but you may greatly reduce the effect it has on the overall mix using this high pass filter. Process only this small portion of the audio file, creating a new region in its place without the pronounced thump. You might have to create crossfades on either side of this newly created region in order to smooth the transition between the original recording and the newly processed one.

This same type of process can be achieved through the use of automation. That technique will be explained in Chapter 8, "Mixing to Picture in Pro Tools". While both methods can potentially achieve the same results, by using the Audiosuite plug-in you can affect very small portions of audio easily. Using the automation method might not yield the same results with very small audio segments due to priorities in the automation structure of Pro Tools and how certain plug-ins react to automation data. In addition, the Audiosuite method frees up DSP power for other tasks , as the file is processed offline.

Once you've completed any the offline processing and all editing tasks are complete, it is time to start mixing.