Singing Voice Synthesis¶
Audio
Open Source¶
-
FM-Singer Discovered by AI
2026
A flow-matching based singing voice synthesis model designed to mitigate the latent mismatch in conditional Variational Autoencoder (cVAE) architectures, enhancing expressive details in synthesized vocals.
cvaeflow-matchingsvs -
SoulX-Singer Discovered by AI
2026
A state-of-the-art zero-shot singing voice synthesis model designed to clone and synthesize singing voices from short reference audio clips. It achieves high-fidelity vocal generation with precise pitch control and expressive style transfer.
svsvoice-cloningzero-shot -
VocalRender Discovered by AI
2026
A score-native singing voice synthesis system that directly generates expressive singing from lyrics, MIDI pitches, note values, and tempo without requiring predefined phoneme-level durations or time-aligned acoustic guidance.
diffusionscore-nativesinging-synthesissvs -
YingMusic-Singer Discovered by AI
2026
A fully diffusion-based model for zero-shot singing voice synthesis and editing. It enables melody-controllable singing generation with flexible lyric manipulation and annotation-free melody guidance, allowing users to alter lyrics while preserving melody consistency.
diffusionsvszero-shot -
DiTSinger Discovered by AI
2025
A Diffusion Transformer (DiT) based singing voice synthesis model that scales SVS with a large-scale corpus of over 500 hours of high-quality Chinese singing data. It incorporates Rotary Position Embedding (RoPE) and implicit alignment to achieve expressive vocal generation.
diffusion-transformerditscalingsvs -
TCSinger 2 Discovered by AI
2025
A multi-task multilingual zero-shot singing voice synthesis (SVS) model that supports style transfer and style control based on natural language, speech, or singing prompts. It uses a Flow-based Custom Transformer and a Mixture of Experts (MoE) architecture to achieve high-fidelity, customizable singing voice generation.
multilingualstyle-transferzero-shot -
TechSinger Discovered by AI
2025
An advanced multilingual singing voice synthesis system supporting five languages and seven vocal techniques, leveraging a flow-matching-based generative model for highly expressive and controllable vocal generation.
flow-matchingmultilingualsinging-voicesvs -
VersBand Discovered by AI
2025
A multi-task song generation framework that integrates lyric generation, melody generation, singing voice synthesis, and accompaniment generation with prompt-based control.
music-generationsinging-voice-synthesissong-generation -
YingMusic-Singer Discovered by AI
2025
A melody-driven zero-shot singing voice synthesis and editing framework based on a Diffusion Transformer (DiT) architecture, eliminating the need for phoneme-level alignment and manual melody annotations.
diffusionsinging-voicesvszero-shot -
GTSinger Discovered by AI
2024
A global multi-technique singing corpus with realistic music scores and phoneme-level annotations of six singing techniques, designed to support various singing voice synthesis and conversion tasks.
datasetsinging-corpussinging-voice-synthesis -
Prompt-Singer Discovered by AI
2024
A controllable singing voice synthesis system that allows users to guide vocal characteristics, emotions, and styles using natural language prompts. It bridges text-based descriptions with acoustic features for intuitive vocal production.
prompt-controlsvstext-to-singing -
StyleSinger Discovered by AI
2024
A singing voice synthesis framework designed for out-of-domain style transfer, enabling the generation of expressive vocals in target styles with limited reference data. It models fine-grained style attributes to capture subtle vocal nuances.
style-transfersvszero-shot -
TCSinger Discovered by AI
2024
A zero-shot singing voice synthesis framework designed to generate high-quality singing voices with unseen timbres and styles from audio and text prompts, utilizing multi-level style control.
style-controlstyle-transfersvszero-shot -
CoMoSpeech Discovered by AI
2023
A speech and singing voice synthesis system based on consistency models. By distilling a multi-step diffusion teacher model into a single-step student model, it achieves high-fidelity audio generation with extremely fast inference speeds.
consistency-modelreal-timesvstts -
Fish Diffusion Discovered by AI
2023
An easy-to-use, open-source framework for text-to-speech, singing voice synthesis, and singing voice conversion based on diffusion models.
diffusionsvcsvs -
So-VITS-SVC Discovered by AI
2023
A highly popular open-source singing voice conversion framework based on SoftVC and VITS, enabling high-fidelity voice cloning and timbre transfer for vocals.
svcvitsvoice-cloning -
StyleSinger Discovered by AI
2023
An out-of-domain singing voice synthesis framework designed for style transfer, utilizing a residual style extractor and a style-sensitive pitch predictor to capture and transfer expressive singing styles from reference audio.
singing-voicestyle-transfersvszero-shot -
NNSVS Discovered by AI
2022
An open-source neural network-based singing voice synthesis library designed for research, supporting multi-stream models, autoregressive F0 models, and neural vocoders.
deep-learningnnsvssvs -
VISinger 2 Discovered by AI
2022
An end-to-end singing voice synthesis system that integrates a digital signal processing (DSP) synthesizer into a variational inference framework with adversarial learning. It directly generates high-fidelity audio waveforms from musical scores without intermediate acoustic feature conversion.
dspend-to-endsvs -
DiffSinger Discovered by AI
2021
An open-source singing voice synthesis (SVS) system based on a shallow diffusion mechanism that models acoustic features conditioned on musical scores. It addresses the over-smoothing problem in traditional acoustic models to generate high-fidelity singing voices.
diffusionsinging-voicesvs -
OpenUtau Discovered by AI
2020
An open-source singing synthesis platform and modern successor to UTAU, featuring a highly customizable editing environment, support for multiple voicebank formats, and advanced phonemizers.
svsutauvocal-synthesis
Proprietary / Industry Tools¶
-
ACE Studio Discovered by AI
2023
A commercial AI-powered singing voice synthesis workstation that generates studio-quality vocals from MIDI and lyrics with precise control over pitch, vibrato, and emotion.
ai-vocalscommercialsvs -
VoiSona Discovered by AI
2022
An AI-based singing voice synthesis software developed by Techno-Speech, offering high-quality, realistic vocal generation with VSTi compatibility for DAW integration.
commercialsvsvst -
CeVIO AI Discovered by AI
2021
A next-generation voice synthesis platform utilizing deep learning to realistically replicate human singing and speaking voices with rich emotional expression.
commercialdeep-learningsvs -
NEUTRINO Discovered by AI
2020
A freeware neural network-based singing voice synthesizer that generates highly natural singing voices from MusicXML score files. It is widely used in the anime and indie music scenes in Japan, featuring popular voice libraries like Tohoku Kiritan.
freewareneural-networkneutrino -
Synthesizer V Studio Discovered by AI
2020
A professional-grade singing voice synthesis software developed by Dreamtonics. It combines a deep neural network-based synthesis engine with a sample-based hybrid approach, offering highly realistic vocals, multi-lingual synthesis, and an intuitive editor interface.
commercialdreamtonicseditorsvs -
Emvoice Discovered by AI
2019
A next-generation vocal synthesis plugin (VST/AU/AAX) that allows music producers to create realistic singing voices by inputting lyrics and melody notes directly within their DAW.
daw-integrationvocal-pluginvst -
VOCALOID Discovered by AI
2003
Yamaha's industry-standard singing voice synthesis software, utilizing advanced AI engines in its latest versions to produce natural, highly expressive vocal performances.
commercialsvsvocaloid -
A commercial singing voice synthesizer with expressive control, a leading tool for anime-style vocal and song production.
commercialsinging-synthesis
