How the MIDI DLS Standard Unified Sound Banks

The Downloadable Sounds (DLS) standard was introduced by the MIDI Manufacturers Association to solve the problem of inconsistent audio playback across different computer hardware. Before its introduction, General MIDI guaranteed which instruments played, but not how they actually sounded, while custom sound formats remained locked to proprietary hardware. DLS created an open, standardized specification that bundled both digital audio samples and synthesis articulation parameters into a single cross-platform file, ensuring that MIDI compositions sounded identical regardless of the playback device or operating system.

The Problem of Proprietary Sound Banks

During the 1990s, General MIDI (GM) provided a standardized list of 128 instrument patch names and program numbers. However, GM only dictated which sound was triggered (such as an acoustic grand piano on Program 1), leaving the sound generation method entirely to the hardware vendor. A composition played on an inexpensive FM synthesis sound card sounded radically different from the same track played on a high-end wavetable synthesis card.

To offer higher-fidelity and custom sounds, hardware vendors developed proprietary formats for loading custom digital samples into synthesizer memory. Creative Labs used the SoundFont (.sf2) format for its Sound Blaster series, while Advanced Gravis used patch files (.pat) for the Gravis UltraSound. Because these formats were tied to specific hardware architectures, game developers and digital composers could not distribute custom sounds without targeting a specific sound card manufacturer.

The Introduction of DLS-1

To break this hardware lock-in, the MIDI Manufacturers Association (MMA) and the Interactive Audio Special Interest Group (IASIG) published the Downloadable Sounds Level 1 (DLS-1) specification in 1997. DLS was built on Microsoft's Resource Interchange File Format (RIFF), saving sound banks under the .dls file extension.

DLS-1 unified sound delivery by addressing two distinct requirements:

  1. WAV Sample Transport: It stored raw linear PCM audio sample data inside the file so that playback devices did not have to rely on their own internal ROM samples.
  2. Standardized Articulation: It defined how playback engines must process those samples. This included parameters for key splits, loop points, pitch envelopes, and low-frequency oscillators (LFOs).

By formalizing the synthesis model, DLS ensured that any compliant software or hardware synthesizer would process parameters like attack, decay, sustain, and release identically.

Expanding Capabilities with DLS-2

While DLS-1 established a common baseline, it lacked the advanced synthesis features required by professional musicians and game sound designers, particularly when compared to mature formats like SoundFont 2. In response, the MMA ratified DLS Level 2 (DLS-2) in 1999.

DLS-2 introduced significant technical improvements to the specification:

Integration and Lasting Impact

The primary driver of DLS adoption was its integration into operating systems and multimedia APIs, most notably Microsoft’s DirectMusic component within DirectX. Game developers could embed custom .dls sound banks directly within game assets, ensuring that background music and dynamic soundtracks played consistently across any Windows PC, even on low-cost integrated audio chips.

Although proprietary formats like SoundFont remained entrenched in desktop production tools, DLS successfully proved that sample-based synthesis could be decoupled from proprietary sound cards. Its underlying architecture later informed mobile audio standards, such as Mobile DLS, which brought standardized, lightweight sample playback to early mobile phones and embedded multimedia devices.