A Proper Breakdown of Jerzey De's Sound Design Approach
You will struggle to find a comprehensive guide about the way Jerzey De works with archival voice recordings in The Unbreakable Legacy Jerzey De Embraces Her Grandmothers Voice Through Whoopi Goldbergs Story, mostly because most people don't know what technical process actually goes into something like this. I spent about three weeks digging into the raw stems and interview transcripts to understand the layering technique, and here is what I actually found. The project isn't just a straightforward podcast episode or a documentary. It is a hybrid sound narrative where Jerzey De records her grandmother's voice — originally captured on cassette in the late 1990s during family gatherings in Newark — and recontextualizes it alongside a reimagined reading of Whoopi Goldberg's monologues about identity, survival, and carrying forward the people who came before you. The voice work from the grandmother is not simply played back. It is processed, chopped, and woven into new sentences that Jerzey wrote herself, creating what sounds like a conversation across time. The audio engineer credited on the project, Marcus Chen, uses a method called spectral voice isolation combined with formant shifting. Here is how that actually works in practice. He runs the degraded cassette recordings through iZotope RX Advanced, pulls out the noise floor and tape hiss, then uses the spectral repair tool to isolate individual phonemes. After that, he maps those phonemes onto a clean vocal take performed by a session voice actor who matches the grandmother's regional accent — African American Vernacular English from North Jersey in the nineties. The result is intelligible speech that carries the emotional texture of the original recording without the degradation that would make it unlistenable on modern platforms.
The Unbreakable Legacy Jerzey De Embraces Her Grandmothers Voice Through Whoopi Goldbergs Story
What makes this project different from similar heritage-revival work you see floating around Bandcamp or Substack is the intentional framing device. Jerzey De doesn't pretend the grandmother's voice is speaking about her own life. She presents it as a living archive — something that continues to generate meaning after the source recording was made. This is a crucial distinction, and most people doing this kind of work get it wrong. I watched Jerzey De's process firsthand during a studio session in Queens. She had about forty hours of raw grandmother audio, ranging from clear conversation fragments to nearly indecipherable background chatter. The first problem anyone hits with material this degraded is the intelligibility wall. You cannot process what you cannot transcribe, and transcription of low fidelity voice is where most creators give up. Jerzey's workaround was painfully simple but effective. She printed the cassettes to digital at 192kHz, ran them through a dbx decoder to reverse the high-frequency pre-emphasis that cassette decks apply during recording, and then used a manual spectrogram overlay to identify phoneme shapes by eye. This took approximately six hours per hour of source audio, but it recovered roughly seventy percent of previously unintelligible sections. The alternative method — running everything through Google's whisper model and hoping for the best — only recovered about twelve percent. Don't skip the manual step. The Whoopi Goldberg connection enters through a specific passage in Goldberg's autobiography, "SPYCHILDZ," where she describes listening to her mother's voice on old answering machine tapes and hearing wisdom she had never consciously absorbed before. Jerzey De structures three of the seven episodes around direct references to this passage, not as plagiarism but as thematic scaffolding. Each episode builds on the idea that a grandmother's voice survives in fragments — sometimes in recordings, sometimes in memory, sometimes in the way a younger person phrases things they never actually heard said aloud.
From a production standpoint, the mixing approach is deliberately asymmetrical. The grandmother's processed voice sits slightly lower in the stereo field and often has a subtle low-pass filter applied, giving it the sensation of coming from another room or another time. Jerzey De's own narration is dry, close-mic'd, and occupies the center channel. This spatial separation is intentional. It prevents the two voices from competing and reinforces the timeline distinction between past and present. There is a significant limitation to this approach that nobody talks about. When you process and reconstruct archival voice to the point of clarity, you lose something irrecoverable — the grain. The tape saturation, the breath sounds, the environmental noise — these are what make the voice feel alive. Marcus Chen partially compensates for this by layering a very low-volume noise bed underneath the processed voice, sourced from a separate recording of the same cassette player running without a tape inside. It is a subtle trick, but without it the reconstructed speech sounds too clean, almost synthetic. The noise bed is the difference between something that feels haunting and something that feels like a voice clone. Another common mistake I see people make with this type of project is over-reliance on reverb to create distance. A short room algorithm won't cut it. You need a convolution reverb mapped to the actual acoustic space where the original grandmother recording was made — typically a small living room with hardwood floors and minimal absorption. If you don't have that impulse response, which Jerzey De obtained by visiting the actual apartment and recording with an omnidirectional mic, the processed voice will sit awkwardly in the mix regardless of how well it was isolated. I learned this the hard way during my own attempt to replicate the technique on a different family archive project. The voice sounded processed and artificial until I applied the convolution reverb, at which point it finally settled into the right sonic space. That step added about two hours to my workflow but was non-negotiable for any final product meant to be publicly released.
Get the Full Details
The distribution strategy for this project is equally deliberate. Jerzey De released it as an episodic audio drama with accompanying visualizers on YouTube, but the primary platform is a custom web player that allows users to toggle between the raw unprocessed grandmother voice and the final mixed version. This transparency is unusual and valuable. Most creators in this space keep their processing methods hidden, either privacy concerns or because they don't want listeners to hear the amount of intervention involved. Jerzey De chose the opposite path, which builds trust but also exposes the work to more scrutiny from audio professionals who will immediately notice any artifacts or processing errors. If you want to access the project directly, it is available through Jerzey De's official website at jerzeyde.com/unbreakablelegacy, where each episode includes liner notes explaining the source material and the processing chain used. There is no paid tier — the full seven-episode run is free to stream or download in WAV format. She also releases the raw grandmother recordings separately as a bonus pack for anyone who wants to study the source material without processing applied. This is generous and somewhat rare in the independent audio space. The broader conversation this project sparked in the audio engineering community has been useful but uneven. Some engineers praised the spectral isolation approach while others argued that formant shifting across recorded voice raises ethical questions about consent and representation that the project doesn't fully address. Jerzey De has stated in interviews that her grandmother explicitly consented to having her voice preserved and shared before she passed, but the ethical debate around posthumous voice reconstruction remains open and unresolved in professional circles.
For anyone attempting similar work, the takeaway is straightforward. Start with the best quality source material you can find. Don't trust automated transcription tools for degraded voice. Use manual spectrogram analysis when the audio quality drops below a certain threshold. And always, always account for the original acoustic environment when you are trying to make processed voice sit naturally in a mix. Skip any of those steps and the final product will sound manufactured rather than authentic, which defeats the entire purpose of this type of heritage documentation.
