Cthulhu Voice Changer: Cosmic Horror DM Guide
Running Call of Cthulhu online puts every voice in your head through a microphone and a VoIP codec. The whispered threat of an unseen entity, the grinding resonance of a Deep One elder, the flat alien cadence of a Mi-Go representative — these require more than a deep chest voice and good acting. This guide covers the full audio architecture for a cosmic horror DM voice mod workflow: presets, platform routing, and how to maintain persona consistency across every NPC in a session.
H.P. Lovecraft’s fiction, now fully in the public domain, describes entities so alien that human language fails to capture them. The tabletop adaptation by Chaosium translates that fiction into dice rolls and investigation mechanics — but the sound of your NPCs is entirely yours to construct. Done well, a processed voice makes the horror tactile in a way that text descriptions cannot.
TL;DR
- Cosmic horror voices need more than pitch shift — combine formant lowering, distortion, chorus/ring mod, and cavernous reverb for genuinely alien results.
- Save each major NPC (Deep One, Mi-Go, Outer God emissary, human cultist) as a named preset with a dedicated hotkey.
- Route via low-latency audio capture virtual device — Roll20, Foundry VTT, and Discord all receive processed audio with no extra config.
- Sub-300ms latency keeps voice switching imperceptible to players during scene transitions.
- Persona consistency across an entire session requires labeled presets, not manual re-tuning.
Why Cosmic Horror Demands a Different Audio Approach
Standard fantasy TTRPG voices — gruff dwarves, hissing villains, pompous nobles — can be handled with pitch and a bit of acting. Cosmic horror is categorically different. Lovecraftian entities are defined by their incomprehensibility. Their voices should suggest that the listener’s auditory system is processing something it was not designed to handle.
That effect doesn’t come from a lower pitch alone. It comes from a combination of qualities that signal “not human”:
- Formant irregularity — vowel shapes that don’t map to any language the players know
- Sub-harmonic rumble — a frequency layer that vibrates rather than speaks
- Wet, reverberant space — suggesting a location that doesn’t obey normal acoustics
- Rhythmic irregularity — delivered through your own pacing, amplified by slight pitch automation
A voice changer with a proper effect chain can construct all of these except the last, which is on you as the performer.
The Four Archetypes: Building Your NPC Voice Library
Every Call of Cthulhu session has a recurring set of voice archetypes. Building presets for each before your session starts is faster and more reliable than improvising mid-encounter.
1. The Human Cultist
The cultist is still recognizably human — but wrong. A slight downward formant shift (-10 to -15%) with a subtle high-frequency cut (low-pass around 6 kHz) and a small plate reverb tail produces a voice that sounds like it has been in the wrong room for too long. Too much processing reads as “monster”; too little reads as “NPC.” Aim for unsettling rather than inhuman at this stage.
2. The Deep One Elder
Deep One voices draw directly from the “deep-ocean entity” archetype: heavy pitch drop (-8 to -12 semitones), strong formant lowering (-25 to -35%), significant sub-octave layer, and a long cavernous reverb (decay 2+ seconds). Add a slow wet chorus or gentle phaser to create the impression of a voice that doesn’t propagate through air in the expected way.
3. The Mi-Go Emissary
Mi-Go communicate through a biological mechanism that Lovecraft described as producing sounds no human throat can replicate. A ring modulator effect at a low frequency (around 80-120 Hz) creates a buzzing, insectoid quality. Combine with moderate pitch shift (-5 to -7 semitones), no reverb (Mi-Go sound precise, not cavernous), and a gentle bitcrusher for a clinical, alien precision.
4. The Outer God Emissary / Dreaming Entity
Azathoth, Nyarlathotep, Shub-Niggurath — these voices should feel like they are being heard inside the skull rather than through the ears. Extreme formant shift, pitch modulation (slight vibrato with a very slow LFO), massive reverb on a parallel send (keep 40% dry signal for intelligibility), and a long pre-delay (50-80ms) that separates each word from its own reverb tail. The dissociation between direct signal and reverb creates a dreamlike split presence.
Effect Chain Settings: Cosmic Horror Recipes
These are starting points, not fixed formulas. Tune to your own voice and microphone.
| Entity Type | Pitch Shift | Formant | Distortion | Reverb Decay | Special |
|---|---|---|---|---|---|
| Human Cultist | -2 semitones | -12% | None | 0.6 s plate | LPF @ 6 kHz |
| Deep One Elder | -10 semitones | -30% | Tube sat 20% | 2.2 s cave | Sub-octave -14 dB |
| Mi-Go Emissary | -6 semitones | -10% | Bitcrush light | None | Ring mod 100 Hz |
| Outer God Voice | -8 semitones | -40% | None | 4.0 s hall | Slow vibrato LFO |
| Investigator NPC | 0 | 0% | None | 0.3 s room | Your natural voice |
Keep a “Narrator” bypass preset — your unprocessed voice — for scene narration and direct player address. The contrast between processed NPC voices and your clean narrator voice reinforces the divide between story-world and story-outside.
low-latency audio capture Routing: Getting Your Voice into Roll20, Foundry VTT, and Discord
The routing model is the same across all three platforms. A voice changer with low-latency audio capture support creates a virtual microphone device that Windows registers normally. Any application that accepts microphone input — Roll20’s browser tab, Foundry VTT’s WebRTC voice, Discord’s voice channel — reads from this virtual device and receives your processed audio.
Step-by-step setup:
- Open VoxBooster, select your physical microphone as input, and confirm the virtual output device is active.
- In Discord: Settings → Voice & Video → Input Device — select the VoxBooster virtual microphone.
- In Foundry VTT: Configuration → Audio/Video → Microphone — same virtual device selection.
- In Roll20: the platform uses your browser’s WebRTC input. Set the virtual microphone as your default Windows input device, or change it per-tab in Chrome’s site settings (chrome://settings/content/microphone).
- Test with Discord’s “Let’s Check” voice test or Foundry VTT’s local test mode before your session starts.
The processed audio route is: physical mic → voice changer → low-latency audio capture virtual device → VoIP app/platform → players. No secondary software, no loopback cables.
Managing Persona Consistency Across a Full Session
A four-hour Call of Cthulhu session can involve 15 to 25 distinct NPC voices. Maintaining persona consistency — so players can recognize the Deep One Elder in act three by the same voice they heard in act one — requires a system, not improvised memory.
Naming conventions that work:
- Use the character name, not a descriptor: “Zadok Allen” not “Old Fisherman”
- Prefix recurring entities: “ENTITY — Deep One Elder”, “ENTITY — Dreaming Voice”
- Mark scene-locked NPCs: “(Act 2 only) — Cultist Informant”
Hotkey mapping:
Assign hotkeys to your most-used presets (typically 4-6 per session) and leave the rest accessible via click. Attempting to keyboard-shortcut 20 presets produces more errors than it prevents. Hotkeys for: Narrator (bypass), your two main antagonists, and one cultist archetype cover 80% of a session’s switching needs.
The reset habit:
At the end of each scene, return to your Narrator preset. This prevents accidentally opening the next scene in a Deep One voice because you forgot to switch back after the previous encounter.
AI Voice Cloning for Recurring Characters
For a campaign with a major recurring entity — the Dreaming Priest who contacts investigators across multiple sessions — consider building a dedicated voice through AI voice cloning. This is particularly valuable for:
- Finale villains who need to sound identical from their first mention to their final confrontation
- Dreamscape voices that recur across multiple sessions as recurring elements
- Entities described in found texts whose “voice” players have been imagining for several sessions before hearing it
AI cloning operates at the phoneme level — it preserves your timing and inflection while converting the full timbral character of the voice, producing a more organic result than standard effect chains alone.
VoIP Codec Considerations: What Survives Compression
Discord and most online TTRPG platforms apply Opus audio compression at 16-32 kbps for voice. This compression is optimized for speech and aggressively cuts low-frequency content below 80-100 Hz and high-frequency content above 7-8 kHz.
Practical implications for your presets:
- Sub-octave layers are partially attenuated by Opus. Increase their gain by +3 to +5 dB compared to offline mixing to compensate.
- High-frequency bitcrushing artifacts (Mi-Go preset) may not survive well. Keep the bitcrush subtle or use ring modulation instead, which sits in the mid-range where Opus preserves fidelity.
- Long reverb tails compress efficiently but may sound slightly muddy. Keep reverb wet mix at 20-25% maximum for VoIP delivery.
- Ring modulation at 80-120 Hz sits exactly at the boundary of Opus’s low-frequency rolloff. Test at 150-200 Hz if the effect sounds thin after encoding.
Always do a Discord voice test with your presets in your actual session setup — headphones, session distance from mic, Opus encoding — rather than testing through studio monitors with lossless audio.
Preparing Your Audio Before a Session
Technical failures during a horror TTRPG session break immersion more severely than in other genres — the sustained dread that Call of Cthulhu builds over two to three hours can evaporate instantly from a five-second audio glitch. A pre-session checklist prevents the most common failures.
15 minutes before session:
- Open voice changer before Discord/Roll20, not after. low-latency audio capture initialization order matters.
- Cycle through all active presets and speak two sentences each.
- Check reverb tails aren’t bleeding into each other at preset boundaries.
- Verify virtual microphone is selected in your VoIP app — not the physical mic.
- Set a limiter on your output chain to prevent clipping from distortion effects at loud input volumes.
Download and Setup
VoxBooster runs on Windows 10 and Windows 11. No kernel driver installation, no system restart required. The free trial includes full processing capability — build all your presets and test your full cosmic horror workflow before committing to a subscription at $6.99/month.
Try VoxBooster free and check the full guide to AI voice changers for games for additional real-time voice setup tips.
FAQ
What is a Cthulhu voice changer? A real-time audio tool that transforms your mic into eldritch, inhuman voices for Call of Cthulhu and other cosmic horror TTRPGs — using pitch shift, formant manipulation, distortion, and reverb to produce Old One whispers, Deep One growls, and alien cadences.
How do I make my voice sound like an Old One in real time? Combine extreme pitch shift (-10 to -14 semitones), heavy formant lowering, a wet chorus or ring-modulator effect, and long cavernous reverb. Route through low-latency audio capture so Roll20, Foundry VTT, and Discord all receive the effect live.
Can I switch between NPC voices mid-session? Yes. Save each NPC as a named preset with a hotkey. Processing delay under 300ms makes switching imperceptible. One keystroke moves you from human investigator to eldritch entity without any audible gap.
Does this work with Foundry VTT and Roll20? Both platforms use your system microphone for WebRTC voice. Select the low-latency audio capture virtual device in Discord, Foundry VTT, or your browser’s microphone settings — no additional plugins required.
Do I need a kernel driver? No. low-latency audio capture injection works at the Windows audio API level — no kernel driver, no elevated UAC prompts, no anti-cheat conflicts on Windows 10 or 11.
How many presets can I realistically manage? 6-10 active presets per session is practical. Organize by character name, prefix recurring entities, and keep hotkeys for your 4-6 most-used voices.
Is H.P. Lovecraft’s work in the public domain? Yes. Lovecraft died in 1937 and his core fiction — Cthulhu, Mi-Go, Nyarlathotep, the full Mythos — has been in the public domain for decades. The Call of Cthulhu RPG is a separate commercial product by Chaosium.