fabian s. klinke

Composing for Endel

notes.

Composing for Endel

Composing for Endel means preparing sounds and rules for a range of possible realizations. A fixed master may follow, but it does not show how the app behaves at runtime. The best public evidence comes from Endel’s artist interviews, project pages, credits and accounts of its sound-design work. Endel has not published its full current authoring interface or production manual.

[!note] Evidence status Reviewed 20 July 2026. The workflow combines documented steps across several projects, which may use different production processes.

Function as the brief

The team begins with a function such as focus, sleep, relaxation or movement. That function shapes musical form and the test for success: a passage can sound strong in isolation and still fail if it attracts too much attention, interrupts continuity or conflicts with the intended state.

Endel’s account of its Sleep soundscape describes a designed macro-form with an introduction, a nature phase, colored noise and a longer overnight phase. The intended use therefore organizes time at the level of phases rather than acting as a label placed on finished ambient music.

For Move, Endel says it first made a demonstration, explained the engine to Miguel and requested specific kinds of musical elements. The artist received a system-facing brief and supplied material that could operate inside an existing behavioral framework.

Documented production pattern

  1. Define the use, intended state, duration, foreground level and broad behavior. Endel may also prepare a demonstration so the artist can hear how material behaves in the engine.
  2. Supply modular assets such as isolated stems, voices, sounds or nearly static elements.
  3. Integrate the material. Producers, sound designers, developers and product staff process the assets and connect them to the system described in Endel sound generation.
  4. Test realizations across different states, then remove or revise material that breaks the function or aesthetic.
  5. Release the bounded system. The app produces individual sessions from the materials, transitions, mappings and limits set by the team.

Endel’s current workflow for streaming releases follows a different path. The artist chooses a function and sends stems, which Endel’s in-house producers turn into a fixed wellness-oriented release. This documents the stem-based production relation. A streaming album does not retain the app’s live adaptation.

Grimes and AI Lullaby

The 2020 AI Lullaby project is a concrete example, not a model for the whole market. Official credits identify Grimes as the source of original vocals and music and say Endel Pacific processed the material. The credited team also includes a sound director, sound designer, CTO, developers, product staff, data analysis, QA and support.

In Endel’s interview, Grimes said it was easy to create stems for the team to “shuffle about” and described the result as live rather than a fixed recording. She also argued that working with technology means working with the people who made the tools. The project shows distributed labor without treating “AI” as an autonomous musician.

In a Vogue interview, Grimes said the initial output was weaker, so she revised her material to suit the system. Dazed reports her account that sharp bell sounds upset her child and prompted another revision. These accounts show iterative audition and correction without disclosing the internal technical change.

Grimes supplied a series of stems, heard generated results, gave feedback and revised material. Endel’s role was to process and integrate the assets. Grimes described the AI as learning, though public evidence does not show whether a model updated, staff changed system rules, she changed the stems or several of these occurred.

Do not claim that Grimes sent “batches” of stems, that Endel “encoded” them through a documented tagging scheme or that an identified model learned directly from her feedback. None of those details is established in the reviewed sources.

The official AI Lullaby video shows the project’s vocal material, timbre, density and aesthetic identity. As a fixed video, it cannot show live adaptation, sensor mapping or whether two sessions differ.

Creative roles

  • The artist produces timbre, harmony, rhythm, voice and other material that can withstand recombination.
  • The sound director and producer translate the function into musical constraints, select material and decide what remains fixed or variable.
  • The sound designer prepares assets and processing behavior for variation across states.
  • The engineer and developer implement mappings, transitions, scheduling, rendering and limits.

Service roles

  • Product and research staff define modes, interface choices, state models and evaluation criteria.
  • QA staff and listeners search possible outputs for abrupt, distracting, repetitive or invalid results.
  • The listener selects the function and may supply context or body data that conditions a realization.
  • The company controls the engine, interface, service access and data relation.

System-facing musicianship requires modules that remain coherent in changing combinations, a clear split between invariants and variables, long-term behavior, the removal of material that fails and repeated evaluation across possible outputs.

The labor is distributed but not equal. Listener data may affect what sounds now, while artists and production teams define the aesthetic possibility space and Endel retains platform control. Public sources do not disclose legal authorship, rights ownership, royalty divisions, payment terms or who can reuse the integrated system outside the service.

The case connects to adaptive, generative and functional music because automation changes when and where people make musical decisions.

Sources

Endel project sources

Interviews and media

14 paragraphs1,011 words7,001 characters