Ir para o conteúdo
Atlas de IA

Geração de música e efeitos

Gerar música ou efeito sonoro a partir de uma descrição

Fala e músicaIniciante #30
entradaTextoÁudio

O texto completo é apresentado em inglês; o título e o resumo estão traduzidos.

O QUE ESTA CAPACIDADE SIGNIFICA

Takes a text description, optionally with lyrics or a style reference, and outputs audio — a full song, an ambient bed or a sound effect. It must handle melody, harmony, arrangement and timbre at once, adding a layer of musical structure on top of speech synthesis. Unlike text-to-speech the output is not language but rhythmically and tonally organised audio.

Como é feita tecnicamente

The mainstream uses an audio latent representation with diffusion or autoregressive generation: audio is first compressed into discrete or continuous tokens, generated under conditioning, and decoded back to a waveform. Structural tags during training teach the model the order of intro, verse and chorus. Lyrics and vocals are produced by separate alignment and synthesis modules and then mixed with the accompaniment into a finished track.

Produtos representativos

3

Organizações relacionadas

Usos típicos

  • Background music for short video and podcasts
  • Sound effects for games and apps
  • Scores for ads and promotional films
  • Creative demos and arrangement ideas

Como avaliar se funciona bem

FAD
Distribution distance between generated and real music; lower is better
MOS
Mean opinion score for listening quality
Human preference
Pairwise judgement of melodic and arrangement appeal

Limites e dificuldades

  • Long-form structure collapses: chorus returns and arrangement layers fail to hold together
  • Lyrics and melody misalign and enunciation blurs, with mispronounced or swallowed syllables
  • Imitating the style of a living artist under copyright carries risk and needs checking before commercial use

Conceitos por trás