Score-native singing voice synthesis system that turns lyrics and a MIDI melody into expressive vocals.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
VocalRender is a singing voice synthesis system that takes a symbolic score (lyrics plus a MIDI melody) and generates sung vocals. According to a YouTube creator review, it combines an autoregressive model with a diffusion model. The project's demo page presents human-reviewed samples that pair each score with ground-truth audio and six system outputs for side-by-side comparison.
The same review describes two open models, each under 10 GB, released with training code. At the time of that review it had been trained only on Chinese, so support for other languages is not confirmed.
The provided pages show only an audio demo site. Vendor, license, repository, model hosting and release date were not found in the material, so those details are unverified.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Singing voice generator that takes lyrics and MIDI melody and outputs expressive vocals, using an autoregressive plus diffusion design. Two open models under 10 GB, with training code; trained only on Chinese so far.
Closest neighbours by meaning, not just by category.
App
Vocal synthesis tool reportedly used by an AI model to generate lead vocals, harmonies and chops for a song.
Model · ElevenLabs
ElevenLabs' text-to-music model for generating songs with vocals, inpainting and genre transitions, cleared for commercial use.
Model
Open-source text-to-music model that generates full songs with vocals from a text prompt and lyrics.
Model · Google
Google DeepMind's music generation model producing vocal tracks up to 3 minutes from text or image prompts.
Model
Open-source diffusion-based music generation foundation model with lyric-aligned vocals and editing controls.
Model · Multimodal Art Projection (M-A-P)
Open-weights music generator that plans an editable ABC-style score, then renders full songs with vocals and accompaniment.
