Re-creating high-fidelity sounds from low-fidelity source?

I suspect the answer is “no”, but I’ll ask: does there exist some reliable technique to process an existing low-fidelity (as in 8-bit, often 22khz or less) sound sample into a higher-quality one that sounds as close to the original as possible, but with none of the nasty lo-fi artifacts? Alternately, does there exist software that can automatically do this?

Or perhaps is this a service that some (most?) professional sound engineers are capable of doing by hand – re-creating a sound from scratch on their own equipment as closely as possible to the “feel” of the original sound?

I ask because I have an absolute mountain of old unreleased tunes I made in FastTracker/ImpulseTracker/ModPlug circa 1998-2002 which I’ve been sitting on all these years. Most are crap, but a handful of them aren’t too bad, compositionally-speaking, and I’ve been pondering the idea of compiling the good ones, polishing them up in Renoise, and releasing them. It’s just that the sound quality of every individual instrument sample in these tunes is poor, so I’m brainstorming what to do about that.

(masking away the most high/low-end artifacts with filters only goes so far…)

Could you share several of these poor samples so I can experiment with them?

Basically, a sample can “suffer” from two things, which are related to the same thing; lack of information:

  1. Lack of information due to low sample resolution
  2. Lack of information in peaks due to excessively high overall amplitude (artifacts).

Artifacts in the second point can be “fixed”. But the missing information in the first point is simply not there. If the sound is “poor”, it sounds bad, and will still sound bad even if you increase the resolution. You can simply equalize/filter to remove part of the sound in the sample that makes it sound bad, for example a particular high or low end. The problem here is that the correct original sound is not available. It is difficult to restore it.

1 Like

This would require VERY big efforts, and even then there’s no guarantee of an authentic “original” sound.

Recreating all of them from scratch would require a lot of experience, a lot of time and a lot of try and error. Even if you’d find someone who could possibly do this, you will probably have to pay a lot. Instead you could also use AI, for example Genopatch that is included in the Synplant 2 VST. First fix everything possible about your low quality samples, then load them into the tool, click the button and you’ll get a recreated synth preset of your sample.

Perhaps you could also load them into Renoise, first fix what can be fixed and then layer everything. So put the original low quality samples on top of a solid high quality foundation. Of course those foundations have to have a sound that’s as close as possible to the original sound or at least complements (“fixes”) the original sample.

But whatever you do, the original sound will get lost either way. You may get something similar, maybe something that’s as close as possible, but not 100% the exact same sound. The character of your songs would change for sure. Personally I don’t think it’s worth the effort restoring tracker tunes from 20+ years ago.

1 Like

From chatgpt;

" You’re right to be cautious about the possibility of automatically turning low-fidelity audio into high-quality sound. There isn’t a magic bullet that can perfectly convert 8-bit, low-sample-rate audio into high-fidelity sound without some trade-offs or significant manual intervention. However, there are some techniques and tools that can help improve the quality of these samples, though they won’t be perfect.

1. Upsampling and Interpolation

  • Upsampling involves increasing the sample rate of the audio. For example, taking a 22kHz sample and converting it to 44.1kHz. While this doesn’t inherently add new information, it can make the audio more manageable for further processing.
  • Interpolation techniques, like spline or sinc interpolation, can smooth out the waveform. This can reduce some artifacts, but it won’t recover lost information.

2. Noise Reduction and Denoising

  • Tools like iZotope RX or Audacity’s noise reduction feature can help reduce some of the noise and artifacts present in low-quality samples. These tools use algorithms to identify and remove unwanted noise, but care is needed to avoid removing desirable elements of the sound.

3. EQ and Filtering

  • Applying equalization (EQ) can help balance out the frequency spectrum, reducing harsh highs or muddy lows. High-pass and low-pass filters can also be used to cut out unwanted frequencies. While this doesn’t add fidelity, it can make the sample more pleasant to listen to.

4. Harmonic Excitation and Synthesis

  • Harmonic exciters or enhancers like those found in some mastering software (e.g., Waves, Ozone) can add harmonic content to the sound, effectively “fattening” it and making it sound richer. This process can sometimes mask the low-fidelity nature of the original sample.
  • Resynthesis is a more complex approach where a synthesizer attempts to recreate the harmonic structure of the original sound. Tools like Celemony Melodyne or Zynaptiq’s Zynaptiq UNMIX::RECONSTRUCT can assist with this.

5. Re-sampling or Re-recording

  • If you have access to higher-quality samples that resemble the original sound, replacing the lo-fi samples with these can be a good solution. This process can be time-consuming but offers the best results.
  • Some sound engineers specialize in re-creating sounds. They can analyze your original samples and recreate them using high-quality synthesizers or instruments, closely matching the original feel but with vastly improved fidelity.

6. AI-based Enhancement

  • There are emerging AI-based tools like Acon Digital’s DeVerberate or iZotope’s Neutron that use machine learning to improve audio quality. These can sometimes remove artifacts or enhance the clarity of sound, but their effectiveness can be hit or miss, especially with extremely low-quality audio.

7. Professional Sound Engineering

  • Professional sound engineers could manually recreate or enhance your samples using a combination of the above techniques. This would involve significant time and expertise but could yield the best results. Engineers might also use advanced analog gear to add warmth and depth to your samples, further improving the quality.

Practical Steps You Could Take:

  1. Test Upsampling and Noise Reduction: Start by upsampling your audio and using noise reduction techniques to clean it up.
  2. Experiment with EQ and Harmonic Enhancement: Apply EQ and harmonic excitation to see if you can improve the perceived quality.
  3. Replace Samples Where Possible: Consider replacing the worst samples with high-quality equivalents if you can find or synthesize them.
  4. Consult a Professional Engineer: If you are serious about the release, you may want to consult a professional who can either enhance or recreate your sounds.

Final Consideration:

While some improvement is possible, it’s essential to manage expectations. The extent to which you can enhance these samples depends on their initial quality and the effort you’re willing to invest. A mix of automated processing and manual work, possibly with the help of a professional, could give your old tracks new life."

Thanks, guys. I’ll definitely give that Synplant thing a try soon, and if I’m still feeling as headstrong/ambitious a few weeks later as I had this weekend regarding the idea of recreating (almost) everything, I’ll commission a pro sound engineer.

“Plan C” is to just scrap all old instrument samples altogether and just use the old tunes’ motifs as basis for composing new tunes from scratch. That’s certainly not a bad option, either.

@Raul I uploaded few samples if you still feel like experimenting: Proton Drive