DiffRhythm is a diffusion-based lyrics-to-song model that creates vocals and accompaniment together. Structured lyrics establish the song, while a style description and optional reference audio guide its musical character. Its official model page describes a ninety-five-second standard mode and a 285-second extended mode, allowing either a shorter demo or a more developed composition. Verse and chorus sections help communicate the intended song layout. Use cases: lyric demos, reference-guided songwriting prototypes and full-song soundtrack sketches built around an existing set of words.
ACE-Step Prompt offers a natural-language workflow for the ACE-Step music foundation model. A brief can describe the scene, genre, mood and intended character of a song in one place, supporting both vocal and instrumental music. It is useful for exploring an idea before writing a separate lyric document; creators with fixed words can use the companion tags-and-lyrics version. Seed and requested-duration controls support repeated exploration of the brief. Use cases: exploratory song concepts, mood-led compositions and instrumental background cues.
ACE-Step is an open music foundation model co-led by ACE Studio and StepFun. Its diffusion-based architecture combines a compressed audio representation with a lightweight linear transformer, with the technical report emphasizing musical coherence and alignment between sung words and music. This version accepts style tags and optional lyrics as separate creative inputs, giving songwriters explicit control over the musical palette and words. Seed and duration controls also support instrumental compositions. Use cases: lyric-led song sketches, genre-tag exploration and music built around a defined song structure.
Stable Audio 3 Medium Base is the foundational 1.4-billion-parameter checkpoint for the family's longer stereo music compositions. It is the unmodified underlying checkpoint in the Medium family. A text brief can establish instruments, atmosphere and an arrangement that develops across the piece, while seed, negative-prompt and duration controls guide generation. Its Base identity is useful when evaluating the underlying model alongside adapted versions. Use cases: baseline long-form composition experiments, cinematic instrumental sketches and comparative studies of musical prompts.
Stable Audio 3 Medium is a 1.4-billion-parameter latent diffusion model for stereo music and longer-form compositions. Its positioning within the Stable Audio 3 family gives arrangements more room to develop themes, changes in texture and a musical resolution. Prompts can describe instrumentation, harmony, mood and the intended progression of a piece, with seed, negative-prompt and requested-duration controls. The official playground illustrates a cinematic piano-and-strings sketch. Use cases: extended instrumental development, cinematic underscore and compositions built around a sustained musical or emotional arc.
Stable Audio 3 Small Music Base is the foundational 459-million-parameter checkpoint in the Small Music family, generating stereo music compositions up to two minutes long from text. The Base designation identifies the underlying, unmodified model checkpoint in this family. For generation, a brief can specify instruments, rhythm and atmosphere, with seed, negative-prompt and duration controls. Use cases: baseline composition experiments, short instrumental beds and comparing how the foundational checkpoint responds to different musical prompts.
Stable Audio 3 Small Music is a 459-million-parameter latent diffusion model for complete stereo music compositions up to two minutes long. It is the compact music-focused member of the Stable Audio 3 family. Text prompts can describe the instrument palette, genre, rhythmic feel and atmosphere, while seed, negative-prompt and duration controls support exploration of a musical idea. Its short composition format is useful when a cue needs a developed arrangement within a modest runtime. Use cases: instrumental cues, layered beat sketches and short-form soundtrack development.
Stable Audio 2.5 is Stability AI's text-to-audio model for music and sound effects. It can turn descriptions of instruments, atmosphere and musical development into evolving audio, with seed and requested-duration controls. The official example develops a quiet piano figure into a larger orchestral moment, illustrating how a prompt can express both the starting material and the intended dynamic arc. This makes it useful for work that combines composition and scene-oriented sound design. Use cases: cinematic underscore, orchestral sketches, evolving musical beds and sound-effect concepts.
Lyria 3.5 is Google DeepMind's music model for complete, structured songs spanning several minutes. Its model card emphasizes musicality, vocal quality and complex arrangements, with natural-language direction for tempo, lyrical timing and the progression of a track. A brief can describe an opening, a change in texture, a breakdown and a final resolution, using timing cues where helpful. This suits creators planning how a composition should unfold over time. Use cases: developed vocal song ideas, hybrid-genre compositions and cinematic music with an emotional arc.
Lyria 3 Pro is Google's full-song model for complete, structured compositions up to three minutes long. A natural-language brief can describe genre, instrumentation, vocal character and a progression such as verse-chorus-verse or a gradual build from gentle to powerful. The official model card also describes lyric timing and tempo direction through the prompt. This gives creators space to develop an arrangement beyond a short musical clip. Use cases: complete song concepts, vocal compositions with planned sections and soundtracks that need a defined musical arc.
Lyria 3 is Google's high-fidelity music model for thirty-second clips, including music with vocals and lyrics. Its model card emphasizes musical awareness and structural coherence within a short composition. A text brief can combine a genre, instrument palette, tempo, mood and vocal direction, allowing a short musical idea to be expressed as a single creative description. This clip-focused format suits quick exploration before developing a longer work. Use cases: vocal-hook ideas, short soundtrack concepts and compact music sketches for content production.
CassetteAI Music Generator creates 44.1 kHz stereo music from a written description and requested track length. A brief can name instruments, key, tempo and mood to turn a general soundtrack idea into a more specific musical direction. Its official example combines piano, bass and drums into a beat-making prompt, illustrating how the arrangement can be described without musical notation. Requested durations range from ten seconds to three minutes. Use cases: instrumental backing tracks, short-form soundtrack cues and musical beds with a defined rhythm and atmosphere.
Sonilo 1.1 generates music from descriptions of style, mood and instrumentation, with an explicit requested duration. Its own model family is developed using licensed music datasets. A prompt can combine musical influences and describe the feel of a backing track rather than being limited to a single genre label. Whole-second duration control, from one second to ten minutes, helps fit a cue to the time available in a creative project. Use cases: mood-led backing music, short-form soundtrack cues and accompaniment matched to a specified runtime.
Mureka 9.5 Instrumental is dedicated to composing music without sung vocals. A natural-language brief can combine instrumentation, tempo, mood and the intended development of the piece, allowing the creator to outline both a sound palette and a musical arc. The official example moves from a sparse opening through a more expansive middle to a gentle resolution. This is useful when music needs to support a scene without competing with dialogue or narration. Use cases: cinematic underscore, acoustic background pieces and instrumental sketches for a defined mood.
Mureka 9.5 Song creates songs with vocals and accompaniment from supplied lyrics and optional musical direction. The brief can describe the singer's character, the instrument palette and how the arrangement develops, while the lyrics keep the intended words explicit. The official example combines an intimate lead vocal with a rhythm section, layered chorus harmonies, an instrumental break and a resolved ending, illustrating a complete song arc. Use cases: lyric-led song demos, vocal arrangement sketches and narrative soundtrack songs with a planned emotional progression.
ElevenLabs Music 2.5 generates music from a detailed natural-language brief. Its official model card describes the revision as building on Music 2 with improved audio quality and prompt adherence. A brief can specify instrumentation, rhythm, musical sections and how the arrangement should develop, from an opening idea through a breakdown to an ending. It supports vocal or instrumental music and requested track lengths from three seconds to ten minutes. Use cases: composed soundtrack cues, instrumental arrangements, backing tracks and music shaped around a detailed creative brief.
ElevenLabs Music 2 is a text-driven music model for vocal and instrumental compositions. The official model card positions it as an improvement over the first Music model in prompt adherence, composition and vocal delivery, with multilingual support. Musical direction can describe a sound palette, groove, melodic hook and progression between sections in one natural-language brief. Requested length and an instrumental-only option help tailor the composition to its intended role. Use cases: song concepts, dance-track sketches, backing music and scene-specific soundtrack cues.
ElevenLabs Music turns natural-language descriptions into vocal or instrumental music. A prompt can establish the genre, mood, instruments and role of the track, from a restrained background bed to a more prominent musical cue. An instrumental-only option allows music without sung vocals, and requested lengths range from three seconds to ten minutes. This makes it useful for describing the atmosphere and musical function of a piece in ordinary language. Use cases: background music, pop-inspired cues, lo-fi accompaniment and soundtracks for a defined scene or runtime.
MiniMax Music 3 is a full-song model for turning written lyrics and musical direction into tracks up to five minutes long. A brief can describe genre, tempo, key, vocal character and instrumentation, while sectioned lyrics establish how the song develops. This makes room for an introduction, verses, a contrasting chorus and an ending within one composition. It is useful for creators who want to shape both what is sung and how the backing arrangement unfolds. Use cases: complete song demos, arranged vocal compositions and lyric-led soundtrack development.
MiniMax Music 2.6 creates complete tracks with singing, backing music and detailed arrangements from lyrics and a style description. The two inputs separate the words from the intended sound: lyrics establish the vocal content, while musical direction describes genre, mood, instrumentation and vocal character. Planning these elements together gives songwriters a way to explore different treatments of a written song idea. Use cases: vocal song sketches, lyric demos and mood-led soundtrack songs.
MiniMax Music 2.5 is a lyric-driven model that turns supplied words and musical direction into a complete vocal track with accompaniment. Its style prompt can describe a genre, emotional atmosphere and the character of the performance, while the separate lyrics supply the actual song. This workflow is useful when the message is already written and the remaining creative task is finding its musical treatment. It generates the song as a composition with vocals and backing instruments. Use cases: songwriter demos, narrative songs and vocal music for content soundtracks.
MiniMax Music 2.0 uses a dual-input composition workflow: musical direction establishes genre, mood and atmosphere, while separate lyrics provide the words and song structure. Section markers such as [Intro], [Verse], [Chorus], [Bridge] and [Outro] can guide how the arrangement progresses. The model produces a complete song with vocals and accompaniment, so the lyrics and intended structure should be prepared before generation. It suits creators who know what they want to say and want to explore how it should sound. Use cases: indie music production, content soundtracks and structured lyric-driven composition.
MiniMax Music 1.5 combines supplied lyrics with a separate description of the desired musical style. It belongs to the earlier MiniMax lyric-driven generation line and is suited to developing a compact written song idea into vocal music. A brief can establish a genre and energy level, while section markers in the lyrics express the intended verse and chorus layout. Keeping words and musical direction separate allows the same idea to be explored with different treatments. Use cases: short song demos, vocal sketches and musical interpretations of an existing text.
MiniMax Music 1 uses an audio reference to guide the musical style of a new vocal composition built around supplied lyrics. The reference provides an existing musical example instead of requiring the creator to express every stylistic detail in writing. The official guide recommends a reference longer than fifteen seconds containing both music and vocals. This makes the model useful for developing a written idea around a sound that is already familiar. Use cases: reference-guided songwriting, short vocal prototypes and exploring new lyrics against an established musical direction.