sona

About & notes

Why these sounds

Sona is a small, quiet instrument: you tell it how you feel, and it plays a room back to you. It is not a playlist and not a diagnosis. It is a way to sit with a feeling for ten minutes, or to ask the feeling to move.

Everything you hear is generated in your browser as you listen. There are no audio files, no accounts, no tracking, and no two listens are identical. The whole site is a few hundred kilobytes of code.

How to use it

What the music is doing

Each room is built from the same small set of musical decisions: tempo, mode, register, attack, density, and the amount of space around the notes. These are the parameters that music-psychology research most reliably links to perceived emotion, even though the link to felt emotion depends on the listener.

Slow tempos and soft attacks tend to lower arousal; minor and modal harmony tends to read as darker or more reflective; wide voicings and long reverb make a sound feel distant and spacious. The rooms avoid sudden loud transients because the brain-stem startle response is one of the fastest ways music can change how you feel — and not always pleasantly.

The more personal mechanisms — a chord that recalls a specific year, a timbre that sounds like a person — cannot be designed for everyone. That is why the rooms stay abstract and instrumental, and why the interface keeps asking you rather than telling you what you feel.

The eight rooms

Three rooms sit at the top because they are the most direct routes to feeling lighter: Joyful (active pleasure), Calm (settling) and Gentle (connection). Inside those rooms they are one tap apart, under “Compare”, so you can hear the difference on the same day instead of guessing from a description.

RoomTempo & keyTextureDesigned for
Joyful120 BPM, C majorBright mallets and plucks, light shaker, staccatoActive lift; movement without shouting
Calm72 BPM, F majorSoft pad, sparse piano, no percussionSettling; steadiness and natural closure
Gentle78 BPM, E♭ majorPlucked nylon-like pattern, warm pad, dry mixCloseness and comfort; company without words
Hope100 BPM, D minor → D majorPiano, strings and light percussion enter over ~50 sEncouragement; the arc from hesitation to confidence
Nostalgia82 BPM, G majorFM bells, vinyl texture, slight rubatoWarm memory, with a caution for low mood
Sadness60 BPM, A minorLow pad, long descending tones, no percussionCompany in sadness; no insistence on cheering up
Tension80 BPM, G phrygianDissonant clusters, irregular entrances, unresolvedUnease and suspense; not for winding down
Anger138 BPM, E minorSaturated synth bass, driving kit, short motifsForce and release; loud on purpose

Tempo and mode are design choices, not prescriptions. The controlled tempo study cited below used a single classical piano excerpt at 56, 106 and 156 BPM with 26 listeners; it is a useful signal, not a rule for all music or all people. The same caveat applies to the eight briefs: they are starting points for experiments.

If you want to make your own

There are three practical routes. They differ in cost, control and how personal the result can be.

1. Generative AI (fastest)

Tools such as MusicGen, Stable Audio, Suno or Udio can produce a usable bed in a few minutes. Ask for instrumental, slow, loopable, and explicitly request no sudden dynamic jumps. The prompt library has a complete 90-second brief for each of the eight emotions; this is the Calm one:

“Compose a 90-second instrumental piece expressing calmness, reassurance, and comfort. Set around 72 BPM with sparse instrumentation using a soft piano and warm sustained strings… Avoid dense arpeggios, strong drum beats, suspenseful harmonies, or sudden volume increases.”

Swap the key, tempo and instruments per room using the table above, or start from the full briefs. Generate three or four candidates, listen critically, and keep the one whose energy matches the mood — not just its label. Check the platform's licence before publishing.

2. A DAW or sample library (most control)

Start with a sustained pad and a sparse melodic element; resist adding a beat until the room needs one. Useful specifications:

3. Real-time synthesis (what this site does)

The rooms are generated with the Web Audio API: oscillators and filtered noise through a shared convolution reverb and a gentle compressor. Each room has a chord progression, a 16-step pattern, and a small set of instrument voices. The scheduler looks 200 ms ahead so timing survives a busy tab.

This approach is cheap, royalty-free and infinitely variable, but it has limits: synthesized strings and pianos are approximations, and a browser cannot match a recorded performance. For a public installation, recorded or AI-generated audio will sound richer; for a prototype or a classroom demo, synthesis is hard to beat.

What to avoid

What the evidence supports — and what it doesn't

The five source documents for this project were checked against publisher metadata and abstracts (Crossref, OpenAlex, Europe PMC). The short version:

The corrections that matter for the argument:

  1. Baltazar & Saarikallio (2016) organise affect self-regulation as Goals – Strategies – Tactics – Mechanisms (GSTM). The third level is tactics, not “specific activities”.
  2. Sachs, Damasio & Habibi (2015) do not use “aesthetic distance” or “catharsis” as their framework. They argue that sad music is pleasurable when it is perceived as non-threatening, is aesthetically pleasing, and brings psychological benefit — thereby correcting a homeostatic imbalance.
  3. Carlson et al. (2015) found that using music to discharge negative emotion was associated with anxiety and neuroticism in all participants, and particularly in males; it is not a male-only effect.
  4. Salimpoor et al. (2011) used PET and fMRI in separate sessions with the same listeners. The caudate/accumbens dissociation is solid; “proves” is too strong a verb.

Two further points shaped the design. First, “music that sounds happy” is not the same as “music that makes you happy”: perceived and felt emotion are different measurements. Second, the iso principle — matching a mood before shifting it — is clinically popular but not clearly superior: a 2026 pre-registered randomised experiment with 193 people found no difference between iso and compensatory playlists, although both guided conditions increased calmness more than unguided listening. That is why Sona offers both a match and a shift, and claims neither is best.

References

All links resolve to publisher records. DOIs were checked against Crossref; open-access records were cross-checked in Europe PMC.

Music, emotion and the brain

Listening, memory and regulation

Music generation and recommendation systems

Limits. Sona is a student project and a design prototype. It has not been evaluated as an intervention. The self-report evidence it draws on is mostly short-term, Western or Chinese samples, and often uses music chosen by the researchers. Nothing here is medical advice. If you are struggling, please talk to someone qualified — and if a piece of music makes you feel worse, stop it. That is allowed.

← Back to the rooms