say I have some old cassettes from the board and just want to toss them on a computer and get some noise off. is there an easy way using SL 10 pro, or a series of steps to take, i have looked at some vids but I am not as bright as the kids today. If you can refer to anything, great, if not, no worries, advice appreciated, going to read the manual now. Thanks
1 Like
Hope this helps!
Here’s a breakdown of how to remove noise from an old cassette recording using Spectralayers 10 Pro. This process involves several tools and techniques, allowing you to address different types of noise present in cassettes.
Understanding Cassette Noise:
Cassette recordings often suffer from a variety of noise problems:
Tape Hiss: A constant broadband noise characteristic of analog tape.
Hum: Low-frequency noise (50Hz/60Hz) often due to electrical interference.
Dropouts: Brief losses of signal, often sounding like clicks or pops.
Wow and Flutter: Variations in tape speed, causing pitch fluctuations. (Spectralayers is less ideal for this, but sometimes can help)
Clicks and Pops: Discrete artifacts caused by dust, scratches, or other imperfections on the tape.
Other Background Noise: Anything else recorded along with the music.
General Workflow (with specific steps for Spectralayers 10 Pro):
Import and Analyze the Audio:
Open the File: Open the cassette recording WAV or AIFF file in Spectralayers 10 Pro. (File > Open)
Initial Inspection: Listen to the recording carefully, paying attention to the types of noise present and their characteristics.
Visual Inspection: Use the spectral display (waveform and spectrogram) to visualize the noise. Look for patterns that differentiate the noise from the desired audio. Zoom in on areas with noticeable noise.
Normalize (if necessary): Consider normalizing the audio to -1dB or -3dB to ensure you’re working with the optimal signal level before noise reduction. (Edit > Normalize)
Noise Profiling and Reduction (Hiss/Broadband Noise):
Find a Clean Noise Section: The most crucial step. Identify a section of the recording without the desired signal (music, vocals) – a silent gap between songs, a very quiet intro, or even just a short, relatively noise-only area. The longer and “cleaner” this section, the better.
Selection: Using the selection tools (e.g., the rectangle selection tool, the freehand selection tool) in Spectralayers, carefully select the noise-only section in both time and frequency. Be precise and avoid selecting any of the desired audio if possible. You can use the “lasso” tools for more complex selection.
Noise Profile: With the noise section selected, go to Process > Noise Reduction > Learn Noise Spectrum. This creates a noise profile based on your selection.
Noise Reduction Application:
Process > Noise Reduction > Noise Reduction. This opens the Noise Reduction dialog.
Adjust Parameters: The key parameters are:
Reduction: Determines the amount of noise reduction applied (dB). Start conservatively (e.g., -6dB to -12dB) and increase gradually. Too much reduction can create artifacts and make the audio sound “dull” or “processed.”
Noise Floor: Sets the threshold below which the noise reduction is applied. Experiment to find the optimal setting.
Frequency Smoothing: Smooths out the noise reduction across the frequency spectrum to avoid unnatural artifacts.
Time Smoothing: Smooths the noise reduction over time.
Preserve Harmonics: Crucially important for music. Enable this if you’re working with musical content. It tries to protect harmonic structures of the music while reducing noise.
Preview: Use the preview function to hear the effect of the noise reduction. Listen carefully for artifacts (e.g., “watery” sound, loss of high frequencies).
Apply: Once you’re satisfied with the preview, click “Apply” to apply the noise reduction to the entire file (or the selected section).
Iterate: Often, a single pass of noise reduction isn’t enough. Repeat the process (selecting a new noise profile, adjusting parameters) for optimal results.
Hum Removal:
Identify the Hum: Hum is typically a distinct tone at 50Hz or 60Hz (and its harmonics). In the spectral display, you’ll see a clear, narrow band.
Use the Frequency Selection Tool: The frequency selection tool allows you to isolate the hum. Draw a narrow band around the frequency of the hum.
Attenuation: Use the amplitude tool (or Edit > Amplitude) to reduce the amplitude of the selected frequency band. Be very careful not to overdo it, as this can affect the surrounding frequencies.
Notch Filter (Alternative): Spectralayers includes notch filters that can be used to target specific frequencies. Try experimenting with Process > Filters > Notch Filter. Set the frequency to 50Hz or 60Hz, adjust the Q factor (bandwidth), and the gain reduction.
Click and Pop Removal:
Visual Identification: Clicks and pops are usually visible in the spectrogram as short, vertical lines or spikes. Zoom in to identify them precisely. Listen carefully to confirm they are indeed clicks or pops.
Selection: Use the selection tools (e.g., the pencil tool, the marquee selection) to carefully select each click or pop artifact without selecting any of the surrounding desired audio.
Repair (best method): Right-click on the selection and choose “Repair”. Spectralayers uses its algorithms to fill the gap smoothly based on the surrounding audio.
Attenuation (Less desirable): As a less preferable alternative, use the amplitude tool to attenuate the selected click or pop significantly (e.g., -20dB or more). However, “Repair” is almost always better because it preserves the surrounding audio.
Dropout Repair:
Dropouts are often larger than clicks and pops, being small gaps in the audio.
Like clicks and pops, select the area of the dropout as accurately as possible.
Use the “Repair” function in Spectralayers to rebuild the missing audio.
For larger dropouts, the results may be less perfect, and manual editing might be required using copy and paste from nearby similar sections if possible.
Wow and Flutter Correction (Limited Possibilities):
Spectralayers is not the ideal tool for wow and flutter correction. Dedicated audio restoration software (e.g., iZotope RX) are better suited.
Spectral Editing (Advanced): If the wow/flutter is very subtle, you might be able to use spectral editing to adjust the frequencies of individual notes or sections that are slightly out of tune due to the speed variations. This is a very advanced technique and requires a good ear and a lot of patience. It’s often best to seek a specialized tool for this task.
Declipper (If Necessary):
If the recording has clipping (distortion from exceeding the maximum audio level), use the declipper tool: Process > Declipper. Experiment with the settings, but be cautious as declipping can sometimes introduce artifacts.
EQ and Final Adjustments:
EQ: Use EQ to further shape the sound of the recording. You might need to boost the high frequencies slightly to compensate for the loss of detail caused by noise reduction. Be subtle.
Compression (Optional): Use compression to even out the dynamic range of the recording, if needed.
Limiting (Optional): Apply a limiter to prevent clipping and maximize the overall loudness.
Saving Your Work:
Save as a new file: Always save your restored audio as a new file to avoid overwriting the original. Use a lossless format like WAV or AIFF.
Tips and Best Practices:
Work Non-Destructively: Always work on a copy of your original file. This allows you to experiment without risking damage to the original recording.
Less is More: Start with subtle adjustments and gradually increase the intensity of the noise reduction or other processes. Over-processing can lead to unnatural-sounding results.
Listen Carefully: Use headphones and listen critically to the audio at each stage of the process. Pay attention to the balance between noise reduction and the preservation of the desired signal.
A/B Comparison: Regularly compare the processed audio to the original to ensure that you’re making improvements.
Patience: Audio restoration is a time-consuming process. Don’t rush it.
Learn Spectralayers: Take the time to learn the ins and outs of Spectralayers 10 Pro. The more familiar you are with the software, the better the results you’ll achieve.
Use a Good Noise Profile: The quality of your noise profile is critical. Make sure it’s representative of the noise you want to remove and that it doesn’t contain any of the desired audio. When possible, try to create multiple noise profiles from different sections of the recording.
Consider a Dedicated Audio Restoration Suite: For very challenging recordings, consider using a dedicated audio restoration suite like iZotope RX, which often provides more advanced tools and algorithms.
By following these steps and tips, you can significantly reduce the noise in your old cassette recordings and bring them back to life. Good luck! Remember that audio restoration is an iterative process, so be prepared to experiment and refine your techniques.
16 Likes
Holy cow, I can’t thanks you enough. I just gave it a quick glance, and couldn’t ask for more. Off to the basement now for noise reduction fun. Thanks again , sir, thanks,.
2 Likes
that"s a complete tutorial in one forum reply! That’s wonderful stuff.
Steve.
2 Likes
Fantastic!
1 Like
I historically simply pre-normalize a transferred file, click sl to knock out any 60hz gremlins, then invoke sl noise reduction, experiment with percentages, erring towards less nr being better, and click to process.
I then secondarily do all the followup restoration on other programs ( eliminate wow/flutter etc).
1 Like
What a great and beautiful run through from BJ_Dobbs!
I’m not sure if all of the parameters of the Noise reduction are still there in that module in SL11 (Denoise), but I could be wrong.
A method I use to take away broad band noise and hiss is with the Unmix Levels module.
The easiest to handle of the 3 different types there is ”Absolute Power” in my opinion. I usually set the Threshold somewhere at -54 to -60 and check with the Prelisten function.
The goal is to put the noise on the resulting Lo level layer so I toggle between the Hi and Lo and watch the Spectrogram to see how much noise will end up on that Lo level layer. I set the color to Inferno, Magma or similar to see better.
And then I listen to the Hi layer to check that it doesn’t sound unnatural.
Back and forth, adjusting the Threshold value.
When happy with the Threshold I hit Apply and that’s it.
If I’m unsure I set the Threshold so that not much of the noise will end up on the Lo level layer. A little goes a long way and it’s easy to overdo it.
And from there one can always do it again (with a new Threshold).
Then I delete the Lo level layer.
In my opinion this Unmix Levels function can clean a file from noise with often less artefacts than the Noise reduction module. I really like this function!
The other Unmix Levels types work as well but have more parameters to understand and perhaps more specialized tasks (like separating two electric guitars on the same layer).
2 Likes
It’s so good that ChatGPT has apparently copied him nearly word for word!! 
2 Likes
I’m just going to add teething to @BJ_Dobbs instruction
as BJ says “visual inspection”
Learning to read the spectrograph does not happen overnight; it takes time and concentrated effort. I’ve recently worked on some of my earlier jobs from last autumn and my workflow has changed considerably. I see things in the spectrograph in a new way after months of working in SL.
Next, “visual inspection”…how should I set my display parameters?
It is very easy to get tripped up here. I, personally, do not cope well with high FFT sizes or very low brightness. 99% of the time for me I have FFT size set to 2048; Resolution and Refinement at max values.
Having an editing strategy is absolutely key; layer coloring and grouping and position while editing are all key to getting good results
1 Like
Just chiming in to say that although Todd here and I do basically the same type of work using mainly SL, and we mostly agree on stuff, =), we do diverge in one point: Refinement Values.
While my dear fellow prefer max values, I happen to find them usually “too much”, and tend to think that lower values like 60% are more useful because it blurs things a bit and doesn’t distract me with a sea of lines. Like, it makes the elements I’m after a bit more discerneable.
Of course I’m sure we both chage those values according to necessity, but just to give another perspective and say to not oversee the power of lower Refinement values. And as a reminder, unlike FFT values (and resolution to a much much lesser extent), Refinement affects visualization only, it doesn’t change any processes or tool behaviour/results.
2 Likes
Nice pinch but, really? Just as many of us use samples in our songs in order to create original sounding ideas, AI
LLMs are quickly being used the same way: not as a replacement but a co-editor.
Remember, it’s not the end product that counts more than the amount of work that goes into creating it, especially when the data already exists and is well known by the contributor.
Attribution is sticky subject when it comes to tutorials and AI. Is available data actually useful in its original format? If one knows all the details as presented in eg my answer, is it fair to insinuate plagiarism to have an AI summarize and format a useful answer in 10 seconds, with accompanying user review and edits, instead of spending an hour looking up original source material and writing it by hand? And when a member needs help straight away, why not help asap if that’s possible without losing one’s own studio time?
I said before in this forum, I frequently use Gemini AI to search and summarize questions I have that are not easily or clearly answered regarding any Steinberg product. There is no difference here except the attribute I wanted to put at the end of my answer was not copied this time, btw. So what?
2 Likes
Easy, killer… I was just kidding around. Sorry it upset you - that wasn’t my intention. Peace out, sir.
2 Likes
I understood @Thor.HOG to say AI was commandeering @BJ_Dobbs excellent post, not the other way around 
3 Likes
@henrique_staino hey man, I hope all is well…I’m such a green newbie compared to you tho…and I’m an old man, so vision getting worse and worse…I need as much clarity as I can get! 
Anyway, I did say the settings I prefer…I just mean users gotta arrive at their own preferences
No problem. I thought I was being trolled for plagiarism, something I take very, very seriously.
Please accept my humble apology and next time I think I’m being trolled, I will be more inclined to ask the person if that’s what they meant to do. My bad.
Peace upon you and much creativity!
2 Likes
No worries, and same here! All good, and have a fabulous weekend!!
1 Like
OK, whew, Thanks to all, even apparently AI. When I posted this, I figured maybe in a couple of weeks someone would post “read the manual” and that would be it. but Jeez, a couple of days and all this? over little ol’ me (and actually, I ain’t that little, but i digress). If we were all in the same location, beers would be on me, because I am still in shock over responses. And even if you still want to chime in with “read the manual,” feel free, i just printed it out and while i am conscious i will do just that. And while my ears may be to old to hear certain frequencies, i am still looking forward to polishing up some of these old tapes, and thanks to all of you , i can now do just that. Thanks again. And free beers for all (offer limited, see details below).
4 Likes
Do I need a cassette player to do this?
at least one cassette tape would also be a good idea
along with electrical power access and some way to monitor the audio
it’s actually a lot more stuff than that, I just thought I’d cover some of the main points
A real pro doesn’t need either of those.
2 Likes