This model recreates Michael Jackson’s voice from the Dangerous era (1991–1993), capturing both his soft, emotional tone and his raspy, powerful delivery heard in live and studio performances of that period.
It was trained in Applio using the RVC V2 framework, with RMVPE for pitch extraction and KLM HiFiGAN for high-fidelity timbre reproduction.

Trained for 500 epochs, the model is optimized to preserve the most iconic vocal traits of MJ’s Dangerous era:

His airy, whisper-like tone in ballads such as Gone Too Soon and Will You Be There.

His gritty, dynamic power in tracks like Give In to Me and Jam.

His natural vibrato, phrasing, and emotional control in mid-tempo songs like Who Is It and In the Closet.

The result is a balanced and expressive voice model capable of adapting to both gentle and high-energy performances, maintaining the warmth, clarity, and passion that defined Michael’s mature sound.

🎧 Technical Details

Framework: RVC V2

Pitch Extractor: RMVPE

Vocoder: KLM HiFiGAN

Sample Rate: 40 kHz

Training: 500 epochs

Software Used: Applio (training and inference interface)

🎶 Dataset (Audio Sources)

The dataset was carefully curated and cleaned from official studio recordings, stems, and high-quality acapellas from the Dangerous era, keeping a balanced mix between soft and raspy vocal passages.

Gone Too Soon

Will You Be There

Heal the World

Who Is It

In the Closet

Keep the Faith

Jam

Why You Wanna Trip on Me

Black or White

Give In to Me

Dangerous

In the Closet (bridge)

All audio was normalized, trimmed, and denoised to ensure consistent quality and clarity across the dataset. https://huggingface.co/JEFFERSONMUSIC/MJDangerousEraV540K/resolve/main/MJDangerousEraV540K.zip?download=true

Additional Details:

Tags: No tags available

Download Link: https://huggingface.co/JEFFERSONMUSIC/MJDangerousEraV540K/resolve/main/MJDangerousEraV540K.zip?download=true Hosted on Hugging Face