This is a product/research proposal, not implemented Premium commands or an approved release contract. Pricing, entitlement, dates and final feature scope remain undecided. The current processing is documented in the other DSP pages and remains distinct from these ideas.
First priorities under consideration#
| Proposal | What it would allow | Work needed before delivery |
|---|---|---|
| Parametric equalizer | Several bell/shelf/high-pass/low-pass filters with frequency, gain, Q and bypass. | Stable filters, safe headroom, presets and smooth changes without clicks. |
| Per-device profiles | Separate headphone/speaker/output settings with import/export and clear active profile. | Profile validation, format/version rules and safe switching. |
| Convolution with user impulses | Apply an FIR filter supplied for a headphone/room response. | Bounded latency, rate policy, safe transitions and response validation. This engine alone is not automatic room correction. |
Q describes how narrow a parametric filter is. An impulse response is a filter's time-domain description; convolution applies it to the music. Such filters need measured/provenanced inputs rather than a claim that any downloaded curve improves sound.
Other research ideas#
Headphone-model EQ depends on measurement provenance, licensing and individual fit. Crossfeed would mix a delayed/filtered portion of each stereo side into the other for headphone listening. A true-peak limiter would add explicit look-ahead latency, ceiling and gain-reduction metering; it is different from the current hard sample guard.
Night-mode compression could reduce loud/quiet contrast for specific listening situations. Optional crossfade could blend unrelated tracks with configurable curves, while preserving continuous albums by default. Channel balance/trims/delay/polarity and bass-management crossovers require suitable output layouts. Measurement-assisted room correction needs calibrated microphones, several positions and acoustic expertise; it is not committed for the first release.
Dynamic EQ/loudness compensation, mid/side width, time stretch/pitch and binaural/spatial rendering have additional perceptual/CPU/data requirements. They remain research candidates. No current CLI/API option activates them. The existing native PCM transport reuses the current DSP; delivering a future effect through it would require explicit settings and validation. Transport itself is not a Premium DSP effect, and original-file Subsonic transfers do not apply server processing.
How future features should be validated#
Keep an explicit bypass; never modify files/tags. Measure response, stability, aliasing, sample/true peaks, loudness scopes, processing joins and block-size invariance. Show actual stage/latency/profile status. Listening preference and fidelity claims require listening validation, not only attractive curves. Hardware joins and NAS budgets still require target testing even for the present foundation.
The DSP product proposal records the detailed research, proposed priorities and source library. It is the reference for future updates; percentages there are planning estimates, not measured quality or availability guarantees.