- Top Pick: Shure MV7+
- Runner-up: HyperX Cloud III Wireless
- Also consider: SteelSeries Arctis Nova 7 Gen 2
Phasmophobia is the co-op horror game that built its identity around your microphone. Four investigators, a haunted building, and a ghost that listens: Kinetic Games designed its evidence gathering around spoken questions, which made Linux players nervous from day one, because the original release leaned on Windows speech services that Proton does not ship. That history still shapes the game’s reputation, and plenty of older guides repeat it as current fact.
Something did change, years ago. In June 2022 Kinetic replaced the voice recognizer inside the game itself, cutting the Windows dependency out entirely. The current status is better than this game’s Linux folklore suggests: the haunting runs, your crew can hear you, and the ghost can understand your questions, as long as one in-game setting points at the right recognizer. This article separates the two voice systems, because they have different answers.
As an Amazon Associate, we earn from qualifying purchases.
Quick answer
Phasmophobia runs well on Linux through Proton, rated Gold on ProtonDB and trending Platinum. Voice chat works, and the ghost can hear you: Kinetic replaced the old Windows-only speech system in 2022 with a cross-platform recognizer, so the Spirit Box and Ouija Board answer on Linux too, provided the in-game recognizer is set to Vosk. Steam Deck is rated Playable.
The status at a glance
There is no native Linux build; Steam lists the game for Windows only, so Proton is the route on desktop and Steam Deck alike. ProtonDB’s current summary, read directly for this article, puts the game at a Gold headline tier, strong confidence, more than seven hundred reports, with best-reported and trending tiers at Platinum. There is no anti-cheat client and no publisher launcher layered on top. What remains is a Unity title with a mature Proton record, plus the voice question this game is famous for.
Valve rates the Steam Deck version Playable. Its published test notes say all functionality is accessible with the default controller configuration and that it performs well at default settings, with deductions for non-Deck controller icons, text that may be hard to read, and single-player requiring an active internet connection. That is a long way from the Deck’s early days, when the game sat at Unsupported with the voice problem at its centre.
Two voice systems, two different answers
Most confusion about this game comes from treating voice as one feature. It is two separate systems, and they never shared the same Linux fate.
The first is voice chat: the proximity and radio talk between players, handled as ordinary transmitted audio. This worked under Proton from early on, because nothing in it needs the operating system to understand your words. Your teammates hear you, and with voice activation on, the ghost reacts to your microphone noise during hunts: audio level, not comprehension.
The second is voice recognition: the game listening for specific phrases so the ghost can respond. This drives the Spirit Box questions, the Ouija Board, and voice input to cursed possessions. Recognition originally ran on Windows speech services, and it is the only part that was ever truly broken on Linux. Keeping them apart matters: a report saying the microphone works is describing chat; one saying the ghost answered is describing recognition.
How the recognition problem actually ended
At launch, recognition ran through the speech tooling built into Windows 10. Under Proton there was nothing underneath that call, so the Spirit Box sat silent no matter how clearly you spoke, even while voice chat worked fine. In early 2022 the Wine project began implementing the missing Windows speech libraries with the open-source Vosk recognizer underneath, work its developers described as driven by this game.
Kinetic then solved it from the other side. A June 2022 update ripped the Windows-specific recognition out of the game and replaced it with a cross-platform system. The developer’s update notes, reported at the time, said the game no longer relied on Windows speech systems, supported many more languages and platforms including Steam Deck and GeForce Now, and needed no setup. The engine underneath is the same Vosk recognizer, which Kinetic later named outright when bringing it to consoles in February 2025.
The practical consequence today is one menu choice. The settings carry a recognizer option with two entries, Windows and Vosk, alongside a voice recognition mode that can switch to text. Leave the recognizer on Windows under Proton and you have recreated the old bug yourself; set it to Vosk and the Spirit Box answers, the Ouija Board responds, and the box’s ghost icon confirms the game heard and understood you. If you would rather not speak at all, text mode, added in December 2022, offers the same questions as selectable prompts, no microphone required.
What current reports add
Recent ProtonDB reports describe a game that runs, with a short list of platform quirks. Several recommend a SteamDeck launch flag on desktop Linux to keep voice chat behaving, and one detailed current report traces silent mid-session deaths to the Steam overlay restarting in a loop, fixed by disabling the overlay and Steam’s game recording. The same report notes PipeWire microphone overruns eased by a pulse latency setting, and warns against the snap build of Steam, whose sandbox interfered with sync handling. None of it concerns recognition.
One forward note: Kinetic has announced the Resonance update for 24 November 2026, moving the game to Unity 6 and expanding voice recognition with commands asking the ghost to interact with nearby objects, plus a Journal page listing usable phrases. Recheck ProtonDB after it lands; engine moves rewrite reports.
Gear for a game you talk to
Recognition quality is microphone quality: the Vosk recognizer can only work with what it hears. These picks cover the desk and the couch.
Shure MV7+
The dedicated-microphone route, for players whose investigations live at a desk. A dynamic capsule rejects off-axis room noise, suiting untreated rooms and loud households, and the hybrid output plugs straight into USB today and an interface later. Its DSP extras are configured in Shure’s desktop app, published for Windows and macOS, so on Linux plan around the onboard touch panel.
Specs:
– Dynamic cardioid capsule with voice isolation design
– Hybrid USB-C and XLR outputs
– Auto Level Mode and digital pop filter
– Real-time denoiser and reverb, set in the Shure app
– LED touch panel for mute, gain and monitoring
– 3.5 mm headphone output for zero-latency monitoring
HyperX Cloud III Wireless
The all-in-one route: headset and microphone in one purchase, with a dongle that Linux treats as ordinary USB audio. The detachable microphone carries an LED mute indicator, useful when a hunt starts and you need certainty about whether you are live. Its DTS spatial processing is Windows software; on Linux you hear the plain stereo mix the game’s positional audio is mixed for anyway.
Specs:
– Angled 53 mm drivers
– Detachable 10 mm microphone with LED mute indicator and pop filter
– 2.4 GHz wireless via USB-C dongle with USB-A adapter
– Battery rated by the maker at up to 120 hours at half volume
– Memory foam cushions and leatherette finish
– DTS Headphone:X spatial audio through Windows software
SteelSeries Arctis Nova 7 Gen 2
The simultaneous-connection pick: its 2.4 GHz link carries the game while Bluetooth keeps a phone or handheld connected, so crew chat and a call never fight over one device. The retractable microphone hides in the ear cup for single-player sessions. SteelSeries’ Sonar software is Windows-only; the headset itself works as standard USB and Bluetooth audio on Linux.
Specs:
– 40 mm neodymium drivers
– Simultaneous 2.4 GHz wireless and Bluetooth 5.3
– ClearCast Gen 2 retractable bidirectional microphone
– Battery rated by the maker at up to 54 hours
– USB-C connection and fast charging
– Works as standard USB audio without vendor software
beyerdynamic DT 770 Pro 80 Ohm
The isolation pick, for solo investigators and anyone pairing a separate microphone with serious headphones. The closed-back cups keep the room out and the whispers in, which is the entire sensory deal of this genre, and the 80 ohm drivers run happily from ordinary outputs and handhelds. There is no microphone at all; that is the point for buyers who already own one.
Specs:
– Closed-back over-ear design for isolation
– 80 ohm impedance, runs from typical headphone outputs
– Frequency response rated by the maker at 5 to 35,000 Hz
– Velour ear pads and padded headband
– Single-sided straight cable, 3 metres
– No microphone; pair with a separate mic for co-op
Valve Steam Deck OLED 512GB
The couch route, and the platform the 2022 recognition fix was partly built for. Valve’s Playable rating covers the full game on the handheld’s own controls, the built-in microphone serves chat and recognition without extra hardware, and the OLED screen’s true blacks suit a game played in dark rooms, in-game and otherwise.
Specs:
– 7.4 inch HDR OLED display, 1280 by 800, up to 90 Hz
– 512 GB NVMe internal storage
– Wi-Fi 6E and Bluetooth 5.3
– Built-in microphone array
– SteamOS with Proton built in
– Sold primarily through Valve’s own store
How to choose
Start with the setting, not the shopping: any working USB microphone serves this game once the recognizer is set to Vosk, so buy for your wider setup. Choose the MV7+ if you record or stream beyond Phasmophobia and want one microphone that outlasts several headsets. Choose the Cloud III Wireless for a single purchase covering chat, comfort and long battery, and the Nova 7 Gen 2 if a second Bluetooth connection matters to your evenings. Choose the DT 770 Pro if you already own a microphone and value isolation over features. Choose the Deck if your investigations happen away from a desk.
Who should skip
Skip the gear spending if your current headset already gives your crew clear voice chat; recognition is software, and it is already fixed. Skip the game itself on any platform if an always-online single-player requirement bothers you, since Valve’s notes flag it plainly. And skip judging this title by pre-2022 guides that describe its voice recognition as Windows-only: that was true once, and repeating it now is the only part of the old problem still circulating.
FAQ
Does the Spirit Box work on Linux now?
Yes. Since Kinetic’s June 2022 update the game uses a cross-platform recognizer instead of Windows speech services. Set the in-game recognizer to Vosk and the Spirit Box and Ouija Board respond to spoken questions under Proton; the Windows entry in the menu is the one that does nothing here.
Is voice chat the same thing as voice recognition in this game?
No. Voice chat is transmitted audio between players and worked on Linux from early on. Voice recognition is the game understanding set phrases for the Spirit Box, Ouija Board and cursed possessions. Only recognition depended on Windows, and only recognition needed the 2022 replacement.
What if I do not want to speak to the ghost at all?
Use text mode. Since December 2022 the game offers a text option for voice interactions, presenting the questions as selectable prompts. It covers the same evidence gathering with no microphone involved and sidesteps every recognition question in this article.
Is Phasmophobia Steam Deck Verified?
No; Valve rates it Playable. The published notes confirm full functionality on the default controller configuration and good performance at default settings, with deductions for controller icons, small text and the internet requirement for single-player.
Final verdict
Phasmophobia’s Linux story is a fix that folklore never caught up with. The game runs at a Gold ProtonDB tier trending Platinum, its voice chat was never the problem, and its voice recognition stopped depending on Windows in June 2022, when Kinetic built the cross-platform recognizer into the game. Set the recognizer to Vosk, disable the overlay if your sessions die silently, and the only thing left to fear is the intended kind.











