Mariah Carey’s New Year’s Eve 2016 didn’t go so well. The pop diva graced a stage in the middle of Times Square as the clock ticked down to 2017 on Dick Clark’s Rockin New Year’s Eve, hosted by Ryan Seacrest. After Carey’s melismatic rendition of “Auld Lang Syne,” the instrumental for “Emotions” kicked in and Carey, instead of singing, informed viewers that she couldn’t hear anything. What followed was five minutes of heartburn. Carey strutted across the stage, hitting all her marks along with her dancers but barely singing. She took a stab at a phrase here and there, mostly on pitch, unable to be sure. And she narrated the whole thing, clearly perturbed to be hung out to dry on such a cold night with millions watching. I imagine if we asked Carey about her producer after the show, we’d get a “I don’t know her.”
These things happen. Ashlee Simpson’s singing career, such as it was, screeched to a halt in 2004 on the stage of Saturday Night Live when the wrong backing track cued. Even Queen Bey herself had to deal with lip syncing outrage after using a backing track at former President Barack Obama’s second inauguration. So the reaction to Carey, replete with schadenfreude and metaphorical pearl-clutching, was unsurprising, if also entirely inane. (The New York Times suggested that Carey forgot the lyrics to “Emotions,” an occurrence that would be slightly more outlandish than if she forgot how to breathe, considering it’s one of her most popular tracks). But yeah, this happens: singers—especially singers in the cold—use backing tracks. I’m not filming a “leave Mariah alone!!” video, but there’s really nothing salacious in this performance. The reason I’m circling around Mariah Carey’s frosty New Year’s Eve performance is because it highlights an idea I’m thinking about—what I’m calling the “produced voice” —as well as some of the details that are a subset of that idea; namely, all voices are produced.
I mean “produced” in a couple of ways. One is the Judith Butler way: voices, like gender (and, importantly, in tandem with gender), are performed and constructed. What does my natural voice sound like? I dunno. AO Roberts underlines this in a 2015 Sounding Out! post: “we’ll never really know how we sound,” but we’ll know that social constructions of gender helped shape that sound. Race, too. And class. Cultural norms makes physical impacts on us, perhaps in the particular curve of our spines as we learn to show raced or gendered deference or dominance, perhaps in the texture of our hands as we perform classed labor, or perhaps in the stress we apply to our vocal cords as we learn to sound in appropriately gendered frequency ranges or at appropriately raced volumes. That cultural norms literally shape our bodies is an important assumption that informs my approach to the “produced voice.” In this sense, the passive construction of my statement “all voices are produced” matters; we may play an active role in vibrating our vocal cords, but there are social and cultural forces that we don’t control acting on the sounds from those vocal cords at the same moment.
Another way I mean that all voices are produced is that all recorded singing voices are shaped by studio production. This can take a few different forms, ranging from obvious to subtle. In the Migos song “T-Shirt,” Quavo’s voice is run through pitch-correction software so that the last word of each line of his verse (ie, the rhyming words: “five,” “five,” “eyes,” “alive”) takes on an obvious robotic quality colloquially known as the AutoTune effect. Quavo (and T-Pain and Kanye and Future and all the other rappers and crooners who have employed this effect over the years) isn’t trying to hide the production of his voice; it’s a behind-the-glass technique, but that glass is transparent. Less obvious is the way a voice like Adele’s is processed. Because Adele’s entire persona is built around the natural power of her voice, any studio production applied to it—like, say, the cavernous reverb and delay on “Hello” —must land in a sweet spot that enhances the perceived naturalness of her voice.
Vocal production can also hinge on how other instruments in a mix are processed. Take Remy Ma’s recent diss of Nicki Minaj, “ShETHER.” “ShETHER”’s instrumental, which is a re-performance of Nas’s “Ether,” draws attention to the lower end of Remy’s voice. “Ether” and “ShETHER” are pitched in identical keys and Nas’s vocals fall in the same range as Remy’s. But the synth that bangs out the looping chord progression in “ShETHER” is slightly brighter than the one on “Ether,” with a metallic, digital high end the original lacks. At the same time, the bass that marks the downbeat of each measure is quieter in “ShETHER” than it is in “Ether.” The overall effect, with less instrumental occupying “ShETHER”’s low frequency range and more digital overtones hanging in the high frequency range, causes Remy Ma’s voice to seem lower, manlier, than Nas’s voice because of the space cleared for her vocals in the mix. The perceived depth of Remy’s produced voice toys with the hypermasculine nature of hip hop beefs, and queers perhaps the most famous diss track in the genre. While engineers apply production effects directly to the vocal tracks of Quavo and Adele to make them sound like a robot or a power diva, the Remy Ma example demonstrates how gender play can be produced through a voice by processing what happens around the vocals.
Let’s return to Times Square last New Year’s Eve to consider the produced voice in a hybrid live/recorded setting. Carey’s first and third songs “Auld Lang Syne” and “We Belong Together”) were entirely back-tracked—meaning the audience could hear a recorded Mariah Carey even if the Mariah Carey moving around on our screen wasn’t producing any (sung) vocals. The second, “Emotions,” had only some background vocals and the ridiculously high notes that young Mariah Carey was known for. So, had the show gone to plan, the audience would’ve heard on-stage Mariah Carey singing along with pre-recorded studio Mariah Carey on the first and third songs, while on-stage Mariah Carey would’ve sung the second song entirely, only passing the mic to a much younger studio version of herself when she needed to hit some notes that her body can’t always, well, produce anymore. And had the show gone to plan, most members of the audience wouldn’t have known the difference between on-stage and pre-recorded Mariah Carey. It would’ve been a seamless production. Since nothing really went to plan (unless, you know, you’re into some level of conspiracy theory that involves self-sabotage for the purpose of trending on Twitter for a while), we were all privy to a component of vocal production—the backing track that aids a live singer—that is often meant to go undetected.
The produced-ness of Mariah Carey’s voice is compelling precisely because of her tremendous singing talent, and this is where we circle back around to Butler. If I were to start in a different place–if I were, in fact, to write something like, “Y’all, you’ll never believe this, but Britney Spears’s singing voice is the result of a good deal of studio intervention”–well, we wouldn’t be dealing with many blown minds from that one, would we? Spears’s career isn’t built around vocal prowess, and she often explores robotic effects that, as with Quavo and other rappers, make the technological intervention on her voice easy to hear. But Mariah Carey belongs to a class of singers—along with Adele, Christina Aguilera, Beyoncé, Ariana Grande—who are perceived to have naturally impressive voices, voices that aren’t produced so much as just sung. The Butler comparison would be to a person who seems to fit quite naturally into a gender category, the constructed nature of that gender performance passing nearly undetected. By focusing on Mariah Carey, I want to highlight that even the most impressive sung voices are produced, and that means that we can not only ask questions about the social and cultural impact of gender, race, class, ability, sexuality, and other norms may have on those voices, but also how any sung voice (from Mariah Carey’s to Quavo’s) is collaboratively produced—by singer, technician, producer, listener—in relation to those same norms.
Being able to ask those questions can get us to some pretty intriguing details. At the end of the third song, “We Belong Together,” she commented “It just don’t get any better” before abandoning the giant white feathers that were framing her onstage. After an awkward pause (during which I imagine Chris Tucker’s “Don’t cut to me!” face), the unflappable Ryan Seacrest noted, “No matter what Mariah does, the crowd absolutely loves it. You can’t go wrong with Ms. Carey, and those hits, those songs, everybody knows.” Everybody knows. We didn’t need to hear Mariah Carey sing “Emotions” that night because we could fill it all in–everybody knows that song. Wayne Marshall has written about listeners’ ability to fill in the low frequencies of songs even when we’re listening on lousy systems—like earbuds or cell phone speakers—that can’t really carry it to our ears. In the moment of technological failure, whether because a listener’s speakers are terrible or a performer’s monitors are, listeners become performers. We heard what was supposed to be there, and we supplied the missing content.
Sound is intimate, a meeting of bodies vibrating in time with one another. Yvon Bonenfant, citing Stephen Connor’s idea of the “vocalic body,” notes this physicality of sound as a “vibratory field” that leaves a vocalizer and “voyages through space. Other people hear it. Other people feel it.” But in the case of “Emotions” on New Year’s Eve, I heard a voice that wasn’t there. It was Mariah Carey’s, her vocalic body sympathetically vibrated into being. The question that catches me here is this: what happens in these moments when a listener takes over as performer? In my case, I played the role of Mariah Carey for a moment. I was on my couch, surrounded by my family, but I felt a little colder, like I was maybe wearing a swimsuit in the middle of Times Square in December, and my heart rate ticked up a bit, like maybe I was kinda panicked about something going wrong, and I heard Mariah Carey’s voice—not, crucially, my voice singing Mariah Carey’s lyrics—singing in my head. I could feel my vocal cords compressing and stretching along with Carey’s voice in my head, as if her voice were coming from my body. Which, in fact it was—just not my throat—as this was a collaborative and intimate production, my body saying, “Hey, Mariah, I got this,” and performing “Emotions” when her body wasn’t.
By stressing the collaborative nature of the produced voice, I don’t intend to arrive at some “I am Mariah” moment that I could poignantly underline by changing my profile picture on Facebook. Rather, I’m thinking of ways someone else’s voice is could lodge itself in other bodies, turning listeners into collaborators too. The produced voice, ultimately, is a way to theorize unlikely combinations of voices and bodies.
Featured image: By all-systems-go at Flickr, CC BY-SA 2.0, via Wikimedia Commons
REWIND! . . .If you liked this post, you may also dig:
Gendered Sonic Violence, from the Waiting Room to the Locker Room-Rebecca Lentjes
Editor’s Note: February may be over, but our forum is still on! Today I bring you installment #5 of Sounding Out!‘s blog forum on gender and voice. Last week Art Blake talked about how his experience shifting his voice from feminine to masculine as a transgender man intersects with his work on John Cage. Before that, Regina Bradley put the soundtrack of Scandal in conversation with race and gender. The week before I talked about what it meant to have people call me, a woman of color, “loud.” That post was preceded by Christine Ehrick‘s selections from her forthcoming book, on the gendered soundscape. We have one more left! Robin James will round out our forum with an analysis of how ideas of what women should sound like have roots in Greek philosophy.
This week Canadian artist and writer AO Roberts takes us into the arena of speech synthesis and makes us wonder about what it means that the voices are so often female. So, lean in, close your eyes, and don’t be afraid of the robots’ voices. –Liana M. Silva, Managing Editor
I used Apple’s SIRI for the first time on an iPhone 4S. After hundreds of miles in a van full of people on a cross-country tour, all of the music had been played and the comedy mp3s entirely depleted. So, like so many first time SIRI users, we killed time by asking questions that went from the obscure to the absurd. Passive, awaiting command, prone to glitches: there was something both comedic and insidious about SIRI as female-gendered program, something that seemed to bind up the technology with stereotypical ideas of femininity.
Speech synthesis is the artificial simulation of the human voice through hardware or software, and SIRI is but one incarnation of the historical chorus of machines speaking what we code to be female. Starting from the early 20th century Voder, to the Cold-War era Silvia and Audrey, up to Amazon’s newly released Echo, researchers have by and large developed these applications as female personae. Each program articulates an individual timbre and character, soothing soft spoken or matter of fact; this is your mother, sister, or lover, here to affirm your interests while reminding you about that missed birthday. She is easy to call up in memory, tones rounded at the edges, like Scarlett Johansson’s smoky conviviality as Samantha in Spike Jonze’s Her, a bodiless purr. Simulated speech articulates a series of assumptions about what neutral articulation is, what a female voice is, and whose voice technology can ventriloquize.
The ways computers hear and speak the human voice are as complex as they are rapidly expanding. But in robotics gender is charted down to actual wavelength, actively policed around 100-150 HZ (male) and 200-250 HZ (female). Now prevalent in entertainment, navigation, law enforcement, surveillance, security, and communications, speech synthesis and recognition hold up an acoustic mirror to the dominant cultures from which they materialize. While they might provide useful tools for everything from time management to self-improvement, they also reinforce cisheteronormative definitions of personhood. Like the binary code that now gives it form, the development of speech recognition separated the entire spectrum of vocal expression into rigid biologically based categories. Ideas of a real voice vs. fake voice, in all their resonances with passing or failing one’s gender performance, have through this process been designed into the technology itself.
A SERIES OF MISERABLE GRUNTS
The first voice to be synthesized was a reed and bellows box invented by Wolfgang Von Kempelen in 1791 and shown off in the courts of the Hapsburg Empire. Von Kempelen had gained renown for his chess-playing Turk, a racist cartoon of an automaton that made waves amongst the nobles until it was revealed that underneath the tabletop was a small man secretly moving the chess player’s limbs. Von Kempelen’s second work, the speaking machine, wowed its audiences thoroughly. The player wheedled and squeezed the contraption, pushing air through its reed larynx to replicate simple words like mama and papa.
Synthesizing the voice has always required some level of making strange, of phonemic abstraction. Bell Laboratories originally developed The Voder, the earliest incarnation of the vocoder, as a cryptographic device for WWII military communications. The machine split the human voice into a spectral representation, fragmenting the source into number of different frequencies that were then recombined into synthetic speech. Noise and unintelligibility shielded the Allies’ phone calls from Nazi interception. The Vocoder’s developer, Ralph Miller, bemoaned the atrocities the machine performed on language, reducing it to a “series of miserable grunts.”
In his history of the The Vocoder, How to Wreck a Nice Beach, Dave Tompkins tells how the apparatus originally took up an entire wall and was played solely by female phone operators, but the pitch of the female voice was said to be too high to be heard by the nascent technology. In fact, when it debuted at the 1939 World’s Fair, only men were chosen to experience the roboticization of their voice. The Voder was, in fact, originally created to only hear pitches in the range of 100-150 HZ, a designed exclusion from the start. So when the Signal Corps of the Army convinced President Eisenhower to call his wife via Voder from North Africa, Miller and the developers panicked for fear she wouldn’t be heard. Entering the Pentagon late at night, Mamie Eisenhower spoke into the telephone and a fragmented version of her words travelled across the Atlantic. Resurfacing in angular vocoded form, her voice urged her husband to come home, and he had no problem hearing her. Instead of giving the developers pause to question their own definitions of gender, this interaction is told as a derisive footnote of in the history of the sound and technology: the punchline being that the first lady’s voice was heard because it was as low as a man’s.
In fall 2014 Amazon launched Echo, their new personal assistant device. Echo is a 12-inch long plain black cone that stands upright on a tabletop, similar in appearance to a telephoto camera lens. Equipped with far field mics, Echo has a female voice, connected to the cloud and always on standby. Users engage Echo with their own chosen ‘wake’ word. The linguistic similarity to a BDSM safe word could have been lost on developers. Although here inverted, the word is used to engage rather than halt action, awakening an instrument that lays dormant awaiting command.
Amazon’s much-parodied promotional video for Echo is narrated by the innocent voice of the youngest daughter in a happy, straight, white, middle-class family. While the son pitches Oedipal jabs at the father for his dubious role as patriarchal translator of technology, each member of the family soon discovers the ways Echo is useful to them. They name it Alexa and move from questions like: “Alexa how many teaspoons in a tablespoon” and “How tall is Mt. Everest?” to commands for dance mixes and cute jokes. Echo enacts a hybrid role as mother, surrogate companion, and nanny of sorts not through any real aspects of labor but through the intangible contribution of information. As a female-voiced oracle in the early pantheon of the Internet of Things, Echo’s use value is squarely placed in the realm of cisheteronormative domestic knowledge production. Gone are the tongue-in-cheek existential questions proffered to SIRI upon its release. The future with Echo is clean, wholesome, and absolutely SFW. But what does it mean for Echo to be accepted into the home, as a female gendered speaking subject?
Concerns over privacy and surveillance quickly followed Echo’s release, alarms mostly sounding over its “always on” function. Amazon banks on the safety and intimacy we culturally associate with the female voice to ease the transition of robots and AI into the home. If the promotional video painted an accurate picture of Echo’s usage, it would appear that Amazon had successfully launched Echo as a bodiless voice over the uncanny valley, the chasm below littered with broken phalanxes of female machines. Masahiro Mori coined the now familiar term uncanny valley in 1970 to describe the dip in empathic response to humanoid robots as they approach realism.
If we listen to the litany of reactions to robot voices through the filters of gender and sexuality it reveals the stark inclines of what we might think of as a queer uncanny valley. Paulina Palmer wrote in The Queer Uncanny about reoccurring tropes in queer film and literature, expanding upon what Freud saw as a prototypical aspect of the uncanny: the doubling and interchanging of the self. In the queer uncanny we see another kind of rift: that between signifier and signified embodied by trans people, the tearing apart of gender from its biological basis. The non-linear algebra of difference posed by queer and trans bodies is akin to the blurring of divisions between human and machine represented by the cyborg. This is the coupling of transphobic and automatonophobic anxieties, defined always in relation to the responses and preoccupations of a white, able bodied, cisgendered male norm. This is the queer uncanny valley. For the synthesized voice to function here, it must ease the chasm, like Echo: sutured by a voice coded as neutral, but premised upon the imagined body of a white, heterosexual, educated middle class woman.
My own voice spans a range that would have dismayed someone like Ralph Miller. I sang tenor in Junior High choir until I was found out for straying, and then warned to stay properly in the realms of alto, but preferably soprano range. Around the same time I saw a late night feature of Audrey Hepburn in My Fair Lady, struggling to lose her crass proletariat inflection. So I, a working class gender ambivalent kid, walked around with books on my head muttering The Rain In Spain Falls Mainly on the Plain for weeks after. I’m generally loud, opinionated and people remember me for my laugh. I have sung in doom metal and grindcore punk bands, using both screeching highs and the growling “cookie monster” vocal technique mostly employed by cismales.
Given my own history of toying with and estrangement from what my voice is supposed to sound like, I was interested to try out a new app on the market, the Exceptional Voice App (EVA ), touted as “The World’s First and Only Transgender Voice Training App.” Functioning as a speech recognition program, EVA analyzes the pitch, respiration, and character of your voice with the stated goal of providing training to sound more like one’s authentic self. Behind EVA is Kathe Perez, a speech pathologist and businesswoman, the developer and provider of code to the circuit. And behind the code is the promise of giving proper form to rough sounds, pitch-perfect prosody, safety, acceptance, and wholeness. Informational and training videos are integrated with tonal mimicry for phrases like hee, haa, and ooh. User progress is rated and logged with options to share goals reached on Twitter and Facebook. Customers can buy EVA for Gals or EVA for Guys. I purchased the app online for my iPhone for $5.97.
My initial EVA training scores informed me I was 22% female; a recurring number I receive in interfaces with identity recognition software. Facial recognition programs consistently rate my face at 22% female. If I smile I tend to get a higher female response than my neutral face, coded and read as male. Technology is caught up in these translations of gender: we socialize women to smile more than men, then write code for machines to recognize a woman in a face that smiles.
As for EVA’s usage, it seems to be a helpful pedagogical tool with more people sharing their positive results and reviews on trans forums every day. With violence against trans people persisting—even increasing—at alarming rates, experienced worst by trans women of color, the way one’s voice is heard and perceived is a real issue of safety. Programs like EVA can be employed to increase ease of mobility throughout the world. However, EVA is also out of reach to many, a classed capitalist venture that tautologically defines and creates users with supply. The context for EVA is the systems of legal, medical, and scientific categories inherited from Foucault’s era of discipline; the predetermined hallucination of normal sexuality, the invention of biological criteria to define the sexes and the pathologization of those outside each box, controlled by systems of biopower.
Despite all these tools we’ll never really know how we sound. It is true that the resonant chamber of our own skull provides us with a different acoustic image of our own voice. We hate to hear our voice recorded because suddenly we catch a sonic glimpse of what other people hear: sharper more angular tones, higher pitch, less warmth. Speech recognition and synthesis work upon the same logic, the shifting away from interiority; a just off the mark approximation. So the question remains what would a gender variant voice synthesis and recognition sound like? How much is reliant upon the technology and how much depends upon individual listeners, their culture, and what they project upon the voice? As markets grow, so too have more internationally accented English dialects been added to computer programs with voice synthesis. Thai, Indian, Arabic and Eastern European English were added to Mac OSX Lion in 2011. Can we hope to soon offer our voices to the industry not as a set of data to be mined into caricatures, but as a way to assist in the opening up in gender definitions? We would be better served to resist the urge to chime in and listen to the field in the same way we suddenly hear our recorded voice played back, with a focus on the sour notes of cold translation.
Featured image: “Golden People love Gold Jewelry Robots” by Flickr user epSos.de, CC BY 2.0
AO Roberts is a Canadian intermedia artist and writer based in Oakland whose work explores gender, technology and embodiment through sound, installation and print. A founding member of Winnipeg’s NGTVSPC feminist artist collective, they have shown their work at galleries and festivals internationally. They have also destroyed their vocal chords, played bass and made terrible sounds in a long line of noise projects and grindcore bands, including VOR, Hoover Death, Kursk and Wolbachia. They hold a BFA from the University of Manitoba and a MFA in Sculpture from California College of the Arts.
REWIND!…If you liked this post, you may also dig:
Hearing Queerly: NBC’s “The Voice”—Karen Tongson
On Sound and Pleasure: Meditations on the Human Voice—Yvon Bonefant
I Been On: BaddieBey and Beyoncé’s Sonic Masculinity—Regina Bradley