Skip to content

🔊 Sound Library Research

Date: 2026-04-23

Purpose: Reference for sound library design decisions. Covers free soundscape sources, competitive category analysis, and clinical evidence for sound therapy approaches.

Part 1: Free Soundscape Sources

Source License Commercial app OK Quality Best for
Freesound.org (CC0 filter) CC0 Yes — unrestricted Up to 96kHz WAV Primary field recording source
Free To Use Sounds Proprietary royalty-free Yes — explicitly permits app dev 24-bit/96kHz WAV Professional loop masters
Pixabay Audio Pixabay License (permissive) Yes — integrated use 44.1kHz MP3 Breadth, gap-filling
OpenGameArt (CC0 packs) CC0 Yes — unrestricted 44.1kHz WAV Curated ambient loops
numpy/scipy synthesis None — generated code Yes — no licence Mathematically perfect Noise colours

How to use Freesound correctly: filter every search to CC0 only (license:cc0). The library contains CC-BY and CC-BY-NC tracks mixed in — the licence is per-sound, not per-library. Only CC0 is safe for commercial app embedding without attribution requirements.

Sources to avoid

Source Reason
BBC Sound Effects (Rewind) RemArc licence — non-commercial only. PSE commercial licence available but not free.
Zapsplat (standard tier) "Primary value" clause in standard licence is a legal risk for a sound therapy app where sounds are a core feature.
Meta AudioCraft / AudioGen CC-BY-NC model licence prohibits commercial use without a Meta commercial licence. AudioGen outputs 16kHz — insufficient quality for a premium health app.
SoundBible Outdated, uneven quality; licence varies per-sound and is unclear for many tracks.

Noise colour synthesis

White, pink, brown, blue, and violet noise are generated programmatically — no sourcing needed. Python with numpy/scipy:

  • White noise: flat power spectrum — np.random.normal()
  • Pink noise (1/f): power decreases 3dB/octave — generated via spectral shaping in FFT domain
  • Brown noise (1/f²): power decreases 6dB/octave — integration of white noise
  • Blue noise: power increases 3dB/octave — high-frequency emphasis
  • Violet noise: power increases 6dB/octave — very bright, clinically relevant for high-frequency tinnitus masking

All exported as 44.1kHz 16-bit WAV, looped seamlessly (no crossfade artefact at loop point).

Part 2: Competitor Sound Category Analysis

Oto

UK tinnitus CBT app. NHS-adjacent, clinically validated (PMC 2025 feasibility study).

  • 98 sounds in 10 categories
  • Confirmed categories: Coloured noise, Urban, ASMR, Nature
  • Differentiator: notch/boost mode on any sound (frequency cutout or boost at tinnitus pitch) — clinically meaningful
  • Average user engagement with sound library: 14 minutes/session

Kalmeda

German DiGA (certified medical device, prescription-required, GKV-covered).

  • Confirmed categories: Wind, Water, Forest, City/Urban
  • Differentiator: marketed as "3D nature sounds" — spatially positioned audio. Only app that explicitly markets spatial audio as a feature.
  • Spatial sound integrated into a 5-level structured CBT programme, not standalone.

ReSound Relief (GN ReSound)

Free app, clinical framing, no hearing aid required.

  • Estimated 50–150 sounds in three categories: Environmental, Music, Therapeutic
  • Confirmed sounds: frogs, birdsong, water, ocean waves, violet noise
  • Differentiator: layer up to 5 sounds simultaneously with individual volume sliders and L/R balance. Most sophisticated mixing in the category.
  • User-tags sounds by function: Soothing / Interesting / Background

Widex Zen

Widex (WS Audiology) standalone app, free.

  • ~10–15 ambient nature sounds: sandy beach, rocky beach, birds by pond, forest wind, tropical night, summer rain
  • Differentiator: Fractal tones — algorithm-generated chime-like tones that never repeat (mathematically infinite variation). Named by colour variant (Aqua, Lavender, etc.). The foundational Widex Zen patent family (US6816599B2 + DE/CA/EP/JP/AU, priority 2000-11-14) is now expired — so a non-Widex generative engine is buildable, subject to a freedom-to-operate review (see fractal-generative-sound-therapy.md). Clinical evidence stays weak: the n=6 pilot (PMC6197965) plus two confounded n=20 studies, none isolating a fractal-tone benefit.
  • Fractal tones are marketed as addressing habituation — users supposedly cannot "tune out" a sound that never repeats. No study isolates non-repetition as the active ingredient; treat this as an unproven design rationale, not an evidence-backed claim.

AudioCardio

Stanford-validated. Fundamentally different approach.

  • No soundscapes at all — uses inaudible/near-threshold tones at personalised frequencies (Threshold Sound Conditioning)
  • Not comparable as a sound library product — different clinical paradigm

Beltone Tinnitus Calmer

  • Essentially ReSound Relief with Beltone branding + tighter hearing aid integration
  • Same layering engine (up to 5 sounds, L/R balance)

myNoise.net

Web-first (iOS/Android app available). Created by audiologist Dr. Stéphane Pigeon. Most technically sophisticated free tinnitus sound tool available.

  • 200+ sound generators, including:
    • Tinnitus Neuromodulator (partnership with Tinnitus Works)
    • White Bursts Noise (patterned, time-varying)
    • Peak Noise Generator (white/pink/brown/grey with frequency presets 60Hz–8kHz)
    • Notch Noise Generator (fully personalised spectral notch)
    • Rain on tent, forest, ocean waves, japanese garden, tropical rainforest, thunderstorm, fireplace, birdsong, binaural beats, singing bowls, and 180+ more
  • Differentiator: 10-band frequency slider on every generator. Allows precise spectral shaping around individual tinnitus frequency.

Generic apps baseline (Noisli, Relax Melodies, White Noise Lite)

Consumer apps establish baseline expectations: 20–50 sounds, nature + noise colours + ambient music, simple layering, no clinical framing.

Part 3: Clinical Evidence by Sound Category

Approach Evidence level Key finding
Notched noise/music Weak/modest — tonal tinnitus only Largest double-blind RCT (Stein 2016, n=100) null on distress; 2 of 3 recent meta-analyses null; 1 barely significant (THI −8.6/3mo, Jiang 2025, n=793). Effective only for tonal tinnitus ≤8 kHz, 1–2 h/day over months. Full dossier: research/notched-sound-therapy.md.
Fractal tones Weak — n=6 pilot + 2 confounded n=20 studies None isolates a fractal-tone benefit. Widex Zen patent family now expired (FTO review still needed).
Broadband noise (white/pink/brown) Standard of care Moderate evidence for loudness reduction and habituation support
Nature sounds Indirect evidence Promotes parasympathetic activation → stress reduction → indirect tinnitus benefit. ATA recommends dynamic, spatially varied sounds for retraining.
Spatial/3D audio Emerging ATA recommends spatially varied nature sounds for cortical retraining. Kalmeda markets this; no other consumer app.
Binaural beats / neuromodulation Emerging Not yet clinical standard

Based on the above research, this is the proposed category structure — spatial audio is the default for every sound in every category, not a separate tier.

Category Sounds (~6 each) Notes
Water Rain (light), Rain (heavy), River, Ocean waves, Waterfall, Lake shore Most popular tinnitus masking category
Nature Forest dawn, Wind through trees, Fire/crackling, Birdsong, Summer night, Tropical rain Relaxation + attention diversion
Noise Colours White, Pink, Brown, Blue, Violet Synthesised — omnidirectional (no spatial position). Violet noise is a differentiator vs. most apps
Urban Calm Coffee shop, Fan/ventilation, Train, Library, Light rain on window Concentration use case; ASMR adjacent
Tone Therapy Singing bowls, Gentle chimes, Soft drone, Tibetan bells, Crystal bowl Sleep + relaxation; matches competitor category

~30 sounds total. Noise colours are the only category without spatial positioning.

Post-MVP differentiator flagged

Notched noise/music is a possible personalization feature, not a flagship therapy. The clinical evidence is weak/modest (the largest double-blind RCT was null on distress; 2 of 3 recent meta-analyses are null), it works only for tonal tinnitus ≤8 kHz, and it requires sustained daily listening over months — so it must be framed as wellness personalization ("tune a sound to your frequency"), never as a treatment claim. It is cheap to build (notch the synthesized noise via the existing FFT pipeline) and free-to-operate, and fills genuine whitespace (the canonical product Tinnitracks was discontinued in 2024). Recommended after the Phase-3 catalog, with an FTO + regulatory copy review first. Full evidence/product/IP analysis + recommendation: research/notched-sound-therapy.md.

Sources