Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsFrank Sinatra did not record Britney Spears’s “Toxic.” In a 2020 experiment, musicians and researchers CJ Carr and Zack Zukowski of DADABOTS used OpenAI’s Jukebox to generate an unstable, Sinatra-oriented approximation from raw audio. The result sounded uncanny because a neural network was attempting to create the singing, lyrics and accompaniment—not because a real Sinatra vocal had been digitally recovered.
What the experiment was
OpenAI announced Jukebox on April 30, 2020, describing it as a research model that generated music directly as audio. Futurism then asked Carr and Zukowski to try an unusual prompt: Britney Spears’s “Toxic” performed in the style of Frank Sinatra. Their report, published in May 2020, presented one of several generated attempts as an extraordinary early example of AI singing. OpenAI’s Jukebox announcement and Futurism’s original account document the episode.
The clip is best described as AI-generated singing conditioned toward Sinatra’s artist representation. It was not a new Sinatra performance, an authenticated archival recording, or a conventional cover assembled from Sinatra vocal stems.
Who made it?
Carr and Zukowski were the musicians and computer-science experts behind DADABOTS, a project focused on algorithmic music. They had worked with systems such as SampleRNN, but told Futurism that Jukebox was better suited to combining the characteristics of different artists and musical settings. DADABOTS operated the model for this challenge; it did not invent Jukebox.
#1 Best Overall
- COMPLETE VOCAL SETUP: shock mount, pop filter and XLR cable included, add an interface and record
- THE SOUND OF HIT RECORDS: the legendary NT1 large-diaphragm condenser voicing trusted in studios for two decades
- WHISPER-QUIET: among the lowest self-noise microphones ever made, nothing between you and the take
- BUILT FOR VOCALS AND STREAMS: tight cardioid pattern focuses on the voice, rejects the room
- IN THE BOX: NT1 Signature (Black), SM6 shock mount with pop filter, XLR cable and dust cover
The practical workflow involved choosing the artist, genre and lyric information, launching generations, listening to the results and repeating the process until a usable passage emerged. The published clip therefore reflects both the model and the researchers’ selection decisions.
How OpenAI Jukebox generated music
Unlike a system that outputs notes, chords or MIDI, Jukebox worked in the raw-audio domain. OpenAI’s paper describes a multi-scale vector-quantized variational autoencoder (VQ-VAE) that compressed audio into discrete codes, followed by autoregressive Transformer models that predicted those codes. The model could be conditioned on an artist, a genre and lyrics, although the lyrics supplied to it were not aligned to every moment of the audio.
That design let the system model qualities symbolic music does not directly contain: vocal timbre, instrumental texture, distortion, room sound and performance artifacts. It also made generation expensive and control difficult. Audio has vastly more information to predict than a sequence of musical notes, so a small error can become an audible problem as a passage continues. See the Jukebox paper, its arXiv record and OpenAI’s technical overview.
Why the singing sounds uncanny
The model was not selecting prerecorded syllables from a Sinatra library. It was predicting a sequence of audio values shaped by learned patterns. Conditioning can push that sequence toward a familiar era, phrasing style or vocal color, but it cannot guarantee a stable identity or accurate pronunciation.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Rank #2
- The price/performance standard in side address studio condenser microphone technology
- Ideal for project/home studio applications
- High SPL handling and wide dynamic range provide unmatched versatility
- Custom engineered low mass diaphragm provides extended frequency response and superior transient response
- Cardioid polar pattern reduces pickup of sounds from the sides and rear, improving isolation of desired sound source.
- Words may be slurred, omitted or repeated.
- Syllables can fall out of time with the melody or accompaniment.
- A voice may become strained, garbled or tonally unstable.
- The apparent singer can change suddenly, including into an unexpected voice or group vocal.
- The melody, instrumentation or song structure may drift into unrelated material.
- A short phrase can sound convincing even when a longer passage collapses.
Futurism’s interview with the creators described repeated lines, lost lyrics, strained passages and abrupt transitions. These are predictable weaknesses of long-form raw-audio prediction, not evidence that a real performer was being manipulated in the studio.
Why one usable result took so much work
In the creators’ account to Futurism, the team generated attempts over roughly a week. A final generation took about a day, and the necessary cloud hardware cost approximately $3 per hour at the time. Those figures describe their 2020 setup and are not a current price quote for modern music tools or the total cost of training Jukebox.
Repeated runs mattered because each generation could diverge dramatically. The researchers had limited ability to edit a mistake, force a complete arrangement or guarantee that the lyrics would remain intelligible. The published result was therefore partly an exercise in waiting for a favorable sample.
Was it really a cover of “Toxic”?
The prompt targeted Britney Spears’s song, but several distinct things must be separated:
Rank #3
- [USB Output] Enables simple setup. USB studio recording microphone kit provides a direct convenient plug-and-play connection to pc and laptop without any additional hardware or drivers for recording vocals, podcasts and Skype. Studio microphone for recording vocals is never been easier to get high-quality sound for your voice and computer-based audio recordings. (Incompatible with Xbox)
- [Excellent Sound Quality] With rugged construction for durable performance, the vocal recording microphone, USB condenser mic for PC,offers a wide frequency response and handles high SPLs with ease. Ideal for project/home-studio applications. The cardioid condenser capsule captures crystal-clear audio from the front and avoid ambient noise when communicating/creating/recording. Comes ready to go with a desktop mic boom arm stand and 8.2ft USB cable, you're guaranteed to get great-sounding results.
- [Durable Arm Set] The podcast microphone bundle with versatile and sturdy broadcast suspension boom scissor arm with 180° up and down rotation, 135° forward and backward extension for optimal adjustment, for capturing your voice in podcast or voiceover. The double pop filter attached on the music recording microphone provides two layers of dissipation, removes the rush of air, minimize the popping sounds or cancel noise that can compromise your recording, great for studio as well as home use.
- [Easy to Attach] The streaming microphone for PC includes adjustable boom studio scissor arm stand that features a heavy-duty combo mount consisting of a sturdy C-clamp and a detachable desktop mount. With 13" fixed horizontal arm and offers a 30" reach, the low-profile, table-hugging design of audio recording microphone allows on-air talent to perform without facial obstruction to record in podcasting or make dubbing sounds for videos, use voice chat in Discord or online conference on Zoom or Skype.
- [The Accessory Package Includes] The studio microphone music recording comes with practical accessories for you to use in most of recording. The scissor arm stand is made out of all steel construction, sturdy and durable, a studio-grade shock mount, a double pop filter, premium 8.2' USB-B to USB-A/C cable, a podcast PC gaming microphone, a user manual and friendly Technical Support.
| Element | What it means here |
|---|---|
| Composition and lyrics | The copyrighted song associated with “Toxic,” including its recognizable words and musical structure. |
| Commercial sound recording | Spears’s released recording, which is not the same audio file as the generated clip. |
| AI output | A newly generated audio file attempting to render the song through learned patterns, with imperfect fidelity. |
| Performance identity | An approximation directed toward Sinatra’s style or vocal characteristics, not an authenticated Sinatra performance. |
Calling the result a newly generated recording does not make it automatically free of restrictions. The model’s output, the underlying song, any recognizable copied elements and the rights associated with a performer’s identity raise separate questions.
The YouTube takedown and what it did—and did not—prove
A later Futurism report said that a related video was removed after copyright claims. That turned an experimental demonstration into an early example of the platform and rights-management problems surrounding generated music. The report does not establish a court ruling on ownership, fair use or infringement, and it does not show that Britney Spears or Sinatra’s estate prevailed in litigation.
A platform claim can remove an upload before a legal dispute is adjudicated. It is evidence of enforcement pressure, not a definitive answer to the legality of the output.
The rights questions are layered
The song and lyrics
Using recognizable lyrics or musical structure can implicate rights in the underlying composition, independently of how the vocal sounds. The U.S. Copyright Office identifies AI-generated material and the use of copyrighted works in AI systems as active policy issues. Relevant materials include the Copyright Office AI initiative and its AI study page.
Rank #4
- 48V Phantom Power Required – The Maono PM320S professional XLR condenser microphone requires a stable 48V phantom power supply. Simply connect it to your audio interface, mixer, or preamp via the included XLR cable for plug‑and‑play operation, and it delivers pristine studio‑grade sound instantly. Whether you’re using it as a microphone for PC, a vocal microphone for singing, or a recording microphone for music production, the external power ensures consistent, reliable performance every time.
- Studio‑Grade Audio: Built with a large 16mm condenser capsule, this XLR condenser microphone captures extended frequency response and superior transient detail—ideal for studio recording, music production, and vocal tracking. It handles high SPL with ease, making it a versatile recording microphone for podcasting, streaming, vlogging, home studio, and content creation. Whether you're a singer, podcaster, or streamer, this condenser mic delivers broadcast‑quality sound for any application
- Low‑Noise Design for Crystal‑Clear Capture: The cardioid polar pattern effectively rejects off‑axis ambient noise, making it an excellent podcast microphone XLR for untreated rooms, while the integrated shock mount reduces vibration rumble from stands or desks. A pop filter and foam windscreen further tame plosives and saliva noise—perfect for vocal microphone for singing, streaming mic, or recording mic for voice‑over work—so your audio stays clean, immersive, and professionally polished.
- All‑In‑One Studio Kit: Engineered with rugged all‑metal housing, this condenser XLR microphone withstands daily studio use and accidental drops. The adjustable metal boom arm bracket offers stable positioning and folds neatly for easy portability. From podcast setups to music recording, it's the ideal studio microphone for producers, streamers, and vocalists—all at an incredible value, making it the go‑to podcast mic and recording mic for music, vocals, and online content creation
- Packing List: package includes microphone*1, boom arm*1, metal shock mount*1, pop filter*1, windscreen*1, XLR to XLR cable*1 and User manual*1
The sound recording
An AI output is not automatically the original commercial recording. That distinction does not eliminate every possible claim involving copied audio, the composition, contractual terms or platform rules.
Voice, persona and publicity
A recognizable voice and a sound recording are different legal subjects. Depending on the jurisdiction and circumstances, imitation can raise publicity, unfair-competition, consumer-protection or contractual concerns. It is imprecise to reduce all of those issues to “copyrighting a voice.”
Whether the output can be copyrighted
The Copyright Office has continued examining when AI-assisted or AI-generated material receives protection. Human creative contributions may matter, while material generated without sufficient human authorship may receive limited or no copyright protection. An output’s copyrightability is separate from whether using a protected song or recognizable persona creates liability. A 2025 Copyright Office update is available at this notice.
The Jukebox license
OpenAI released Jukebox’s code and weights for noncommercial use. Its repository and license state that restriction and say OpenAI did not claim ownership of content created with the software. Those terms are not a blanket declaration that every output is legally unrestricted. OpenAI later characterized Jukebox as a noncommercial research project and said it had not undertaken significant further work on it in its Senate questions-for-the-record response.
Best Value
- Features a 1” true condenser capsule that captures every nuance of your performance with an outstanding amount of depth and clarity
- Cardioid polar pattern guarantees effective rear rejection, making the LCT 440 PURE ideal for studio, stage, and home recording applications
- Delivers great results on all vocal and instrument applications - vocals, acoustic instruments, drums, cymbals, and amplifiers, piano
- Shock mount and magnetic pop filter are included
How to judge the clip
The experiment is easier to understand when four tests are kept separate:
- Vocal resemblance: Does the timbre or phrasing evoke Sinatra?
- Lyric fidelity: Does it follow “Toxic” accurately?
- Musical coherence: Does it sustain melody, rhythm, instrumentation and structure?
- Authenticity: Is it clearly labeled as synthetic rather than presented as a genuine unreleased recording?
Jukebox could be impressive on the first test while failing the next two. That is the point of the demonstration: generating a recognizable, bizarre hybrid at all was historically notable, even though it was not a commercially polished performance.
Why this 2020 demo still matters
The clip arrived during an early wave of public interest in generative music, years before today’s consumer-facing music services. It showed four capabilities that later became central to the field: artist or genre conditioning, singing generated directly as audio, lyrics used as a control signal and cross-artist combinations that would be impractical for a conventional recording session.
It also exposed the trade-offs that remain important. Raw-audio realism came with weak editability; artist conditioning could evoke a style without preserving a stable identity; lyric prompts did not ensure intelligibility; and longer passages offered more opportunities for drift. The copyright claim demonstrated a parallel problem: technical novelty does not settle rights in the song, the recording or the performer persona.
Recommended Free Tools
OpenAI’s Jukebox was therefore an early research milestone, not a one-click voice-cloning product. Its achievement was the ability to synthesize a plausible but unstable musical hybrid from learned audio patterns—while making clear how far that was from a reliable, authorized Sinatra recording.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
