Music models
Choose a Music model on the Music node (Музыка). Music is a separate audio asset: choosing a music model does not change the native audio generated by a Video node. Connect an approved track to the later audio or assembly step.
Describe genre, instruments, pace, mood and the purpose of the track. Connect the sung words as text to the лирика (lyrics) input when the chosen model supports vocals. The lyrics panel displays that connected text. Do not rely on a request for “background music” to guarantee an instrumental result on a vocal model.
For a structured MiniMax Music 3 brief, use Music Caption Rewriter; for Eleven Music 2.5, set its скил field to ElevenLabs Music v2.5 to get a compact paragraph that fits the 4,100-character prompt together with the lyrics. Put the musical description in Caption, only sung words and bracketed section/control tags in optional Lyrics, and the scenario or other context in an extra text input. Connect its промпт output to Music’s matching input. Its лирика output is a clean sheet of tags and sung lines, so connect it only for a vocal track. For several voices on Eleven Music 2.5, connect both outputs from the same rewriter set to the ElevenLabs skill: Music then sends a composition plan, one chunk per section, at the same per-minute price. MiniMax Music 3 sings in one voice and does not guarantee a duet. The rewrite does not generate audio or new lyrics and has its own cost. See the rewriter guide for the separate caption check, version updates and review boundaries.
Use an existing track
Section titled “Use an existing track”Connect an audio file or a video to the Music node’s аудио input, select Stable Audio 3 or Stable Audio 2.5, and describe the desired music. A video contributes its soundtrack. The source influences the generated track; listen to the full result before using it.
MiniMax Music 3 and Eleven Music 2.5 do not accept this input in Vini. With audio connected, the model picker shows compatible choices and labels an incompatible current selection. Choose Stable Audio or disconnect the audio before running. Stable Audio remains instrumental: the lyrics input does not enable singing on these models.
| Model | New track from text | Track using source audio |
|---|---|---|
| Stable Audio 3 | $0.0376 per generation | $0.0417 per generation |
| Stable Audio 2.5 | $0.20 per generation | $0.20 per generation |
Rates checked on 16 October 2026 against the connected pricing configuration and fal’s Stable Audio 3 audio-to-audio page and Stable Audio 2.5 audio-to-audio page. A mode change changes the estimate. Transport and billing were tested with mocked model results; these checks do not establish output quality.
To describe an existing recording without creating a track, use Audio Breakdown (Разбор аудио). It accepts audio or video, a question and an optional detailed-analysis setting, then returns the full answer as text. See the node reference.
Stable Audio 3
Section titled “Stable Audio 3”The default for a new Music node. Vini connects instrumental generation for this record. Start with a description of instruments, rhythm and energy, then listen for unwanted vocal-like sounds.
| Setting | Value |
|---|---|
| Model ID | stable_audio_3 |
| Name | Stable Audio 3 |
| Availability | available |
| Default for new nodes | yes |
Common declared limits
Section titled “Common declared limits”| Setting | Value |
|---|---|
| Duration | 1–380 seconds; fallback 30 seconds |
| Vocals | no |
Limits apply together. initial is the studio choice for a new node; default is the fallback for an existing node without that value. Declared language or input support does not guarantee a successful result.
MiniMax Music 3
Section titled “MiniMax Music 3”A vocal-capable choice. Separate the arrangement brief from the exact lyrics. A model-author example involving covers or reference audio does not imply those workflows are connected on this Music node.
| Setting | Value |
|---|---|
| Model ID | minimax_music_3 |
| Name | MiniMax Music 3 |
| Availability | available |
| Default for new nodes | no |
Common declared limits
Section titled “Common declared limits”| Setting | Value |
|---|---|
| Duration | 1–300 seconds; fallback 30 seconds |
| Vocals | yes |
Limits apply together. initial is the studio choice for a new node; default is the fallback for an existing node without that value. Declared language or input support does not guarantee a successful result.
Eleven Music 2.5
Section titled “Eleven Music 2.5”A vocal-capable choice. State the intended genre and structure, and supply lyrics when the words matter. Other composition or editing tools described by the author are outside this node’s exposed contract.
| Setting | Value |
|---|---|
| Model ID | elevenlabs_music_25 |
| Name | ElevenLabs Music v2.5 |
| Availability | available |
| Default for new nodes | no |
Common declared limits
Section titled “Common declared limits”| Setting | Value |
|---|---|
| Duration | 3–600 seconds; fallback 30 seconds |
| Vocals | yes |
Limits apply together. initial is the studio choice for a new node; default is the fallback for an existing node without that value. Declared language or input support does not guarantee a successful result.
Stable Audio 2.5
Section titled “Stable Audio 2.5”The previous default remains offered, and older Music nodes without a model choice retain this legacy default. It is not a retired or hidden model. Vini connects instrumental generation for it.
| Setting | Value |
|---|---|
| Model ID | stable_audio_25 |
| Name | Stable Audio 2.5 (прошлая версия) |
| Availability | available |
| Default for new nodes | no |
Common declared limits
Section titled “Common declared limits”| Setting | Value |
|---|---|
| Duration | 1–190 seconds; fallback 30 seconds |
| Vocals | no |
Limits apply together. initial is the studio choice for a new node; default is the fallback for an existing node without that value. Declared language or input support does not guarantee a successful result.
Studio example
Section titled “Studio example”Choose an instrumental record and enter: “A gentle acoustic passage for a quiet workshop scene: plucked nylon guitar, soft brushed percussion, a steady unhurried pulse, and a simple ending.” Set duration for the intended use, generate, and listen to the complete track before connecting it to assembly.
For a song, select a vocal-capable record, write the music brief separately, and supply a short original lyric such as “The morning train is leaving / A new road starts today.” Check pronunciation and whether the vocal phrase finishes within the selected duration.
Verification and evidence
Section titled “Verification and evidence”These are unrendered authoring examples. This guide verifies the current selector, duration and vocal contracts; it does not provide a comparative listening benchmark. Audition the beginning, transitions and ending, and check that the track does not mask dialogue. See sound design and the MCP execution boundary before automating the workflow.
Reviewed 16 October 2026. Model-author sources: Stable Audio prompting, MiniMax music, and ElevenLabs models. Author-site account availability and tools are separate from Vini’s connected model records. A model or version change requires a new review.