Spectre Voice (Coming Soon)

Nyxen Spectre Voice extends Nyxen’s doctrine to voice: time-limited, encrypted calls with no recordings, no voicemail, and no history. It is designed for conversations that should exist only while people are speaking—not as files on someone else’s infrastructure.
This page defines the intent and behavior of Spectre Voice for users, auditors, and implementers.
Concept
Spectre Voice is:
a temporary encrypted voice session
accessed via Nyxen keys / Capsules
bounded by:
a maximum call duration
optional per-session TTL
strictly no recording, no transcription
designed to be burned along with its parent context
Use cases:
sensitive internal syncs
incident calls
decision/approval moments
negotiation or briefing windows
[!IMPORTANT] Spectre Voice is explicitly designed without server-side recording or transcription. If you need a record, Nyxen is not the right channel.
Core Properties (Planned Design)
Access Model
Linked to Nyxen key / Capsule / Dead Drop
Encryption
End-to-end (DTLS-SRTP / WebRTC-style stack)
Duration Limit
Hard cap per session (e.g. 5–45 minutes)
Session TTL
Optional, e.g. must start within X minutes
Recording
Not provided. No server-side capture.
Identity
Pseudonymous within session; no public handle system
Burn Behavior
Burn context → voice handles invalidated, no rejoin
How Spectre Voice Fits Into Nyxen
Spectre Voice is not a standalone calling app.
It is always anchored to another Nyxen primitive, for example:
A Dead Drop Room with an attached Spectre Voice bridge
An Ephemeral Board session with a “Start Spectre Voice” button
A Capsule that includes:
Dead Drop
Board
File Drops
Signals / Ghost Codes
Spectre Voice session
The pattern: one operation, one environment, one burn event.
Capsule: DEAL-ALPHA
- Dead Drop Room
- Ephemeral Board
- File Drop: draft packet
- Spectre Voice: 20-min window
TTL: 90 minutes
Burn: ends all components, including callSession Lifecycle (Intended Behavior)
Initiate
Negotiate
Talk
End
Burn
1. Initiate
From a supported Nyxen context (e.g. Capsule, Dead Drop):
User clicks “Start Spectre Voice”.
Client:
verifies TTL / policies
derives a voice session key from the existing context key.
No global “call ID” directory; everything is scoped.
2. Negotiate (Connection)
Spectre Voice uses a WebRTC-style flow:
DTLS-SRTP for media encryption
Signaling via Nyxen relay using encrypted channels
const voiceKey = deriveKey(capsuleKey, "spectre-voice");
startEncryptedVoiceSession({ voiceKey, ttlSeconds });All signaling messages are encrypted.
Nyxen relays do not see call content or keys.
3. Talk
During a Spectre Voice call:
Audio is:
encrypted end-to-end in transit
never written to disk server-side
UI:
shows participants as generic peers (e.g. “Peer 1”, “Peer 2”)
shows remaining time
indicates when TTL or Capsule expiry is near
No features:
No hold music
No voicemail
No call logs with identity tagging
No built-in transcripts
[!NOTE] Users may still record locally (OS-level or hardware). Nyxen design does not and cannot prevent endpoint capture.
4. End
Calls terminate when:
someone hangs up, or
max duration reached, or
parent TTL (Dead Drop / Capsule) ends, or
a burn event is triggered.
On termination:
Session keys cleared client-side.
Any ephemeral metadata (call start/end timestamps in context) can be:
disabled by policy, or
kept minimal (e.g. “Spectre Voice session occurred during Capsule lifetime”).
5. Burn
If the parent context is burned:
All active or pending Spectre Voice sessions:
are force-terminated,
cannot be rejoined,
cannot be re-established with the old keys.
onContextBurn(contextId) {
endAllSpectreVoiceSessions(contextId);
clearVoiceKeys(contextId);
}[!WARNING] Burn is absolute. There is no way to “restore a call” or pull a recording from Nyxen.
Recommended Usage Patterns
Incident response bridge
10–20 minutes
Capsule + Board + Signals + Spectre Voice
Legal / negotiation touchpoint
10–15 minutes
Use Board for points; no transcripts
Internal high-sensitivity sync
5–15 minutes
Treat as verbal-only; decisions documented in your own infra if required
Escalation call after Ghost Code
5–10 minutes
E.g. Ghost Code 777 → join Spectre Voice
[!TIP] Think of Spectre Voice as a secure corridor: enter, speak, leave. If something must be recorded for compliance, handle it intentionally and outside Nyxen.
Security Considerations
Spectre Voice aims to mitigate:
server-side call recording risk
logging of call contents
linkable identity/account-based call histories
It does not eliminate:
endpoint compromise (malware, rootkits)
physical environment risks (shoulder-surfing, nearby devices)
user-level recording (screen/audio capture)
Implementation should:
use strong, modern voice encryption (DTLS-SRTP)
enforce:
short, non-configurable maximum durations
auto-termination at context expiry
not expose:
phone numbers
SIP addresses
persistent call IDs across contexts
Integration with Tor & Network Constraints
Spectre Voice over Tor is:
possible but constrained
subject to higher latency and reliability issues
Guidance:
mark Spectre Voice as “network dependent”
recommend:
using it primarily where connectivity supports low-latency encrypted media
allowing Tor users to fall back to Dead Drops + Boards if media is unstable
Status
Spectre Voice is defined here as a design and intent:
“Coming Soon” in UI
this page sets:
expectations for ephemerality,
non-recording posture,
key/TTL alignment with Nyxen doctrine.
As it ships, implementation details (codecs, TURN usage, specific crypto libs) should be added here without changing the core guarantees:
No archives. No server-side recordings. No weakening of Nyxen’s model.
Last updated

