GET

/v1/schemas/elevenlabs/audio/v1/music/detailed/stream

Accept:

kind inputactivity audiohash 901b286d…extracted Nov 5, 2026version currentsource upstream spec ↗

field	recorded source
schema	Source linked above. upstream-spec Source trace `1fbf939a1e97e541ce9b5f19ec2b8c81cd5610fb4b8426c6bb70fc579f32679f` { "fetchedAt": 1793856031, "extractorVersion": "1", "verifiedAt": null }

Request body

15 properties · 0 required · 10 $defs

property	type	description	constraints
prompt	string	A simple text prompt to generate a song from. Cannot be used in conjunction with `composition_plan`.	length ≤ 4100 · nullable
generation_mode	enum	Optional generation mode hint for prompt-based music generation. Can only be used with `prompt`.	track \| loop \| ambience \| video_to_music · nullable
music_prompt	MusicPrompt	Composition plan for the `music_v1` model. Using this field with any other model will result in an error.	nullable · deprecated
lyrics_text	string	The lyrics text to use for the generation.	length ≤ 4000 · nullable
composition_plan	MusicPrompt \| CompositionPlan	A detailed composition plan to guide music generation. Cannot be used in conjunction with `prompt`.	nullable
music_length_ms	integer	The length of the song to generate in milliseconds. Used only in conjunction with `prompt`. Must be between 3000ms and 600000ms. Optional - if not provided, the model will choose a length based on the prompt.	3000 ≤ n ≤ 600000 · nullable
model_id	enum	The model to use for the generation.	music_v1 \| music_v2 \| music_v2_5 · default: "music_v1"
seed	integer	Random seed to initialize the music generation process. Providing the same seed with the same parameters can help achieve more consistent results, but exact reproducibility is not guaranteed and outputs may change across system updates. Cannot be used in conjunction with prompt.	0 ≤ n ≤ 2147483647 · nullable
force_instrumental	boolean	If true, guarantees that the generated song will be instrumental. If false, the song may or may not be instrumental depending on the `prompt`. Can only be used with `prompt`.	default: false
finetune_id	string	The ID of the finetune to use for the generation	length ≤ 100 · nullable
finetune_strength	number	How strongly the finetune influences the generation. Defaults to 1.0 (full strength). Lower values soften the influence of the finetune, leaving more room for prompt-level steering. Only meaningful when `finetune_id` is also provided.	0 < n ≤ 2 · default: 1
use_phonetic_names	boolean	If true, proper names in the prompt will be phonetically spelled in the lyrics for better pronunciation by the music model. The original names will be restored in word timestamps.	default: false
store_for_inpainting	boolean	Whether to store the generated song for inpainting.	default: false
with_timestamps	boolean	Whether to return the timestamps of the words in the generated song.	default: false
with_waveform_visual	boolean	Whether to return the visual waveform of the generated song.	default: false

Types

10 $defs

AudioRefChunkobject

song_id*	string	The ID of the song to source the chunk from. You can find the song ID in the response headers when you generate a song.	length ≥ 1 · length ≤ 100
range*	TimeRange	The time range to extract from the song.

CompositionPlanobject

Composition plan for the `music_v2` and `music_v2_5` models. Using this field with any other model will result in an error.

chunks*

GenerationChunk-Input | AudioRefChunk[]

The chunks that make up the generation.

items ≤ 30

ConditionStrengthenum

low | medium | high | xhigh

GenerationChunk-Inputobject

text*	string	The text config to be generated for this chunk. Can contain an optional section name in square brackets at the beginning, e.g. [Verse 1], lyrics lines, and inline directions in curly braces, e.g. {scratching}. Section names must be between 1 and 100 characters…	length ≤ 6132
duration_ms*	integer	The duration of the chunk in milliseconds. Must be between 3000ms and 120000ms.	3000 ≤ n ≤ 120000
positive_styles*	string[]	The styles and musical directions that should be present in this chunk. Use English language for best results. The styles for the first chunk are the most important as they set the overall tone and genre. Styles for subsequent chunks can be used to add nuance,…	items ≤ 50
negative_styles	string[]	The styles and musical directions that should not be present in this chunk. Use English language for best results. Leaving empty is a good default, only use this field if you want to explicitly avoid a particular style or direction.	items ≤ 50
context_adherence	enum	How much the model adheres to the context of its surrounding chunks. Low adherence means the model can deviate from the context and be more creative. High adherence means the model will be more consistent with the context.	low \| medium \| high · default: "high"
conditioning_ref	AudioRefChunk	The audio reference to condition the generation on. The first chunk is the most important as it will influence the generation of all subsequent chunks. Thus, if you want to apply conditioning to the entire song, start conditioning from the first chunk.	nullable
condition_strength	enum	How strongly the model adheres to the conditioning reference. Low strength means the model will be more creative and deviate from the reference. High strength means the model will be more consistent with the reference.	low \| medium \| high \| xhigh · nullable

MusicGenerationModeenum

track | loop | ambience | video_to_music

MusicModelIDenum

music_v1 | music_v2 | music_v2_5 · default: "music_v1"

MusicPromptobject

Composition plan for the `music_v1` model. Using this field with any other model will result in an error.

positive_global_styles*	string[]	The styles and musical directions that should be present in the entire song. Use English language for best result.	items ≤ 50
negative_global_styles*	string[]	The styles and musical directions that should not be present in the entire song. Use English language for best result.	items ≤ 50
sections*	SongSection[]	The sections of the song.	items ≤ 30

SectionSourceobject

song_id*	string	The ID of the song to source the section from. You can find the song ID in the response headers when you generate a song.	length ≥ 1 · length ≤ 100
range*	TimeRange	The range to extract from the source song.
negative_ranges	TimeRange[]	The ranges to exclude from the 'range'.	items ≤ 10

SongSectionobject

section_name*	string	The name of the section. Must be between 1 and 100 characters.	length ≥ 1 · length ≤ 100
positive_local_styles*	string[]	The styles and musical directions that should be present in this section. Use English language for best result.	items ≤ 50
negative_local_styles*	string[]	The styles and musical directions that should not be present in this section. Use English language for best result.	items ≤ 50
duration_ms*	integer	The duration of the section in milliseconds. Must be between 3000ms and 120000ms.	3000 ≤ n ≤ 120000
lines*	string[]	The lyrics of the section. Max 30 lines per section and max 200 characters per line.	items ≤ 30
source_from	SectionSource	Optional source to extract the section from. Used for inpainting.	nullable

TimeRangeobject

start_ms*	integer
end_ms*	integer

Validate a payload

before you spend tokens

shell

$ curl -X POST https://modelschemas.com/v1/validate \
    -d '{"provider":"elevenlabs","endpointId":"v1/music/detailed/stream","payload":{…}}'
→ {"valid": false, "errors": [{"path": "#", "keyword": "required", …}]}