This publication states figures in decibels and grades subtitle implementations against fixed criteria, which is only defensible if the procedure behind them is public. What follows describes the desk, the measurements and the weighting precisely enough to be argued with.
One pair of open-back headphones and one pair of near-field speakers, both used unprocessed. No spatial enhancement, no equalisation, no virtual surround, no software upmixing of any kind. Every game is played through the same chain so that figures from different reviews can be compared with each other.
Games are played on default audio settings first, in full, and only then are the options explored. A review describes the experience a player receives without intervention, followed by what the options can and cannot repair.
A comfortable level is established using an ordinary spoken passage, chosen from early dialogue rather than a cutscene. That level is then fixed and not touched again for the remainder of the playthrough, because a listener who adjusts the volume continuously cannot report on whether adjustment was necessary.
The dialogue-to-effects gap is the difference in level between the quietest dialogue a player is expected to follow and the loudest incident a player is expected to survive, both measured at that fixed reference. It is reported in decibels to one decimal place and represents the widest gap we encountered, not an average across the game.
Dynamic range is recorded separately, from the softest passage in ordinary play to the loudest, because a narrow gap achieved by compressing everything is not the same accomplishment as a narrow gap achieved by mixing carefully.
Positional accuracy is assessed by identifying a sound source with the screen visible, then repeating the identification with the screen switched off. A game passes when we can locate gameplay-relevant sources by ear alone in enclosed spaces, and we report how many attempts each test required.
Where a game is intentionally disorienting, that intention is respected and noted rather than penalised. A space designed to confuse a listener has succeeded when it confuses them, and our figures distinguish deliberate vagueness from careless imaging.
Four criteria, each pass or fail. Whether text can be enlarged and whether the default is readable at a normal seating distance. Whether contrast holds against the game's brightest scene rather than its menus. Whether speakers are identified, particularly off-screen and radio voices. Whether sounds carrying gameplay information are captioned rather than only speech.
The fourth criterion is weighted most heavily in our judgement, because a game whose design depends on a sound effect has removed that design entirely for a player who cannot hear it.
For strategy and management games, alert frequency is measured against progression stage in interruptions per minute, because level-based figures describe such games poorly. A game that scales alerts by importance rather than by the number of entities capable of producing one is credited for it.
One audio score out of ten to one decimal place. It describes the sound and nothing else. A game may have a superb mix and be tedious to play, or a poor mix and be excellent, and this publication only reports on the former.
| Band | Meaning |
|---|---|
| 9.0–10 | Exemplary mix and imaging, subtitles pass everything, works on both outputs |
| 8.0–8.9 | Strong throughout with one identified compromise |
| 7.0–7.9 | Good material undermined by a wide gap, a repeated technique or weak captions |
| 6.0–6.9 | Depends on equipment the game does not mention, or is fatiguing over a session |
| 5.0–5.9 | Information a player needs is not reliably audible or readable |
| Below 5 | Sound actively obstructs play |
Approximately thirty-five per cent of the score reflects mix balance and dynamic range. Twenty-five per cent covers positional accuracy. Twenty per cent addresses subtitle and caption quality. The remaining twenty per cent concerns options and output flexibility, including whether the game acknowledges any equipment it depends on.
All passes are conducted on the build available at the time, on default settings first, with no mods and no third-party audio processing anywhere in the chain. Playthroughs are complete: a game is not scored on a partial run, and speaker checks cover at least three hours.
Measurement errors are corrected in place with a note, and the original figure stays visible. A patch that changes something we measured, such as a rebalanced mix or newly added effect captions, prompts a re-pass and a revised score with the previous figure retained for comparison. Nothing is revised because taste has moved on.
No advertising, no sponsorship, no affiliate arrangements, no relationship of any kind with publishers, platforms or hardware manufacturers. Games and equipment are bought at our own expense. This site provides no correspondence route and no facility for submitting anything to us.
Music criticism, voice performance assessment and hardware reviews are outside our scope. A composer's work may be mentioned where the mix affects how it is heard, but this desk measures delivery rather than composition, and a thin version of somebody else's speciality would serve nobody.
