Munin Labs — Documentation Portal

v4.9

Munin Articulate™

Updated today

59,459

Training Examples

+640 last 24 hours

99.94%

Uptime (30 days)

SLA target: 99.9%

Active Languages

incl. Northern Sami

PASS

Munin Deploy Gate™

1,500/1,500 tests passed

Available Voices

Nora, Erik + 66 more

Platform Improvements

Munin Huginn™ continuously improves the platform. The chart shows the growth of training data and model quality over time.

Recent Updates

Today

Munin Articulate™ v4.9 deployed — 59,459 training examples, 1,500/1,500 Deploy Gate™ PASS

Today

Genesys TTS Connector (BYOT-A) implemented — ready for integration

Yesterday

Documentation portal, status page and admin panel launched

3 June

New language: Northern Sami (davvisamegiella)

2 June

Munin Deploy Gate™ v4.8 passed: 0 regressions

Key Metrics

Munin Accelerate™ hit rate	94%
Response time with Accelerate™	Sub-2 ms
Time to first byte (streaming)	From <100 ms
Pronunciation corrections (nb)	123 rules
Pronunciation languages supported	13
Stress test passed	63,000+ requests
Latest deploy gate	PASS

Platform Overview

Munin Labs operates a fully isolated infrastructure within AWS EU regions. All processing takes place within the customer's chosen AWS region. Data never leaves the EU.

Production Architecture

Language Endpoints

Each language has a dedicated endpoint. No language detection required — the endpoint enforces the correct language. This eliminates misclassification and delivers the highest possible quality.

HTTPSnb.api.muninlabs.io

HTTPSnn.api.muninlabs.io

HTTPSsv.api.muninlabs.io

HTTPSda.api.muninlabs.io

HTTPSfi.api.muninlabs.io

HTTPSis.api.muninlabs.io

HTTPSsme.api.muninlabs.io

HTTPSen.api.muninlabs.io

HTTPSde.api.muninlabs.io

HTTPSfr.api.muninlabs.io

HTTPSes.api.muninlabs.io

HTTPSit.api.muninlabs.io

HTTPSnl.api.muninlabs.io

HTTPSpl.api.muninlabs.io

HTTPShu.api.muninlabs.io

HTTPSpt.api.muninlabs.io

HTTPSja.api.muninlabs.io

HTTPSar.api.muninlabs.io

Customer-Isolated Endpoints

For enterprise customers we provide fully isolated endpoints with dedicated data storage, individual SLA targets and segregated reporting.

Example: Customer-isolated setup Upon onboarding, the customer's own endpoints are provisioned automatically:


            nb.{customer}.api.muninlabs.io

            sv.{customer}.api.muninlabs.io

            da.{customer}.api.muninlabs.io

Each customer receives isolated Munin Accelerate™, dedicated logging, and full GDPR deletion with a single command.

Capability	Description
Data isolation	Fully separated data storage per customer. No cross-contamination possible
Munin Accelerate™	Dedicated per customer. No sharing of optimised data
GDPR deletion	A single API command permanently and irreversibly deletes all customer data
Per-customer SLA	Individual uptime monitoring and reporting
Logging	Segregated logging — each customer sees only their own data

Security Overview

Munin Labs is built with security as a foundational principle. All data processing takes place in AWS EU regions, with encryption at every layer and no third-party services involved in data processing.

Encryption

In transit	TLS 1.3
At rest	AES-256
Certificate management	Automatic renewal
API key hashing	bcrypt (cost 14)
WebSocket	WSS (encrypted)

Access Control

Authentication	API key + HMAC signature
Authorisation	Per-customer isolation (RLS)
Admin access	SSH key pairs, no passwords
Network access	Only port 443 open externally
Internal communication	Private VPC, no public IPs

Data Residency

Guaranteed EU Processing All data processing takes place in AWS EU regions (Ireland/Frankfurt/Stockholm). No data leaves the EU region. No US access. No third-party services in the data processing chain.

Data Type	Storage	Retention
Audio (speech)	RAM only during processing	Never stored
Transcriptions	Encrypted, per-customer isolation	Configurable
Metadata	Encrypted database in EU	Configurable
Model data	Encrypted in EU region	Permanent (not customer data)

Munin Sentinel™ — Continuous Monitoring

Munin Sentinel™ monitors the entire platform in real time, 24/7/365. Anomalies trigger automatic remediation to maintain service quality.

Check	Frequency	Action on anomaly
Service health	Every 60 seconds	Automatic restart within 30 seconds
Response time	Every 60 seconds	Buffer flush + escalation
Capacity	Every 60 seconds	Automatic scale-up within 5 minutes
Security	Continuous	Blocking + alerting
Certificates	Daily	Automatic renewal

Security Policies

Policy	Status	Standard
Information Security Policy	Active	ISO 27001
Acceptable Use	Active	SOC2 CC1
Access Control Policy	Active	SOC2 CC6
Incident Response	Active	SOC2 CC7
Change Management	Active	SOC2 CC8
Risk Management	Active	SOC2 CC3
Privacy Policy	Active	GDPR
Backup Policy	Active	SOC2 CC9

GDPR & Privacy

Munin Labs is designed from the ground up for GDPR compliance. We act as a data processor on behalf of the customer. No personal data leaves the customer's chosen EU region.

Data Flow Diagram

Privacy Guarantees

Audio is never stored	Guaranteed
No third-party services	Guaranteed
EU data residency	Guaranteed
Per-customer data isolation	Guaranteed
Right to erasure	API available
Data Processing Agreement	Available

Compliance Roadmap

Standard	Status
GDPR	Compliant
NIS2	Aligned
DORA (finance/insurance)	Aligned
SOC2 Type I	Planned Q3
ISO 27001	Planned Q4
Schrems II	Not applicable

Schrems II Because all data processing occurs in the EU using EU-based infrastructure, Schrems II concerns are eliminated. No data is transferred to third countries.

Customer Data Deletion (GDPR Art. 17)

Upon termination of a customer relationship or upon request, all customer data is permanently deleted:

Request received

➜

Identity verification

➜

Deletion from all systems

➜

Munin Accelerate™ purged

➜

Confirmation sent

Maximum processing time: 72 hours. Typical: under 24 hours.

SLA — Service Level Agreement

Munin Labs commits to 99.9% uptime for the production environment. In the event of an SLA breach, credits are issued according to the agreed compensation model.

99.9%

Uptime Target

Max 43 min downtime/month

From <100 ms

Streaming TTFB

Time to first byte

Sub-2 ms

Munin Accelerate™

Known requests

Failover Architecture

SLA Tiers

Component	Target	Max downtime/month	Alerting
API availability	99.9%	43 minutes	Immediate
TTS streaming TTFB	From <100 ms	N/A	On threshold breach
TTS with Munin Accelerate™	Sub-2 ms	N/A	On threshold breach
Failover time	<60 seconds	N/A	Automatic
Planned maintenance	Zero-downtime	0 minutes	48-hour advance notice

Incident Response

Severity	Response Time	Resolution Time	Example
Critical (P1)	15 minutes	4 hours	Total service outage
High (P2)	1 hour	8 hours	Degraded performance
Medium (P3)	4 hours	24 hours	Single service failure
Low (P4)	Next business day	5 business days	Cosmetic / minor

SLA Breach Compensation

Uptime	Credit
99.0% — 99.9%	10% of monthly fee
95.0% — 99.0%	25% of monthly fee
<95.0%	50% of monthly fee

Development & Deployment Process

Munin Labs follows a strict development process with automated testing, a staging environment and zero-downtime deployments. No code change reaches production without passing Munin Deploy Gate™.

Deployment Pipeline

Code change (Git)

➜

Munin Deploy Gate™
1,500 automated tests

➜

Deploy to DEV

➜

Manual verification

➜

Promote to PROD
Zero-downtime hot-swap

Munin Deploy Gate™

Every update must pass 1,500 automated tests before it can be deployed. The gate test covers:

Text-to-speech quality	150 x 10 batches
Norwegian (Bokmål)	Tested
Norwegian (Nynorsk)	Tested
Regression	0 allowed
Automatic rollback	On failure

Gate failure blocks deployment If a single test fails, production remains unchanged. No manual override is possible. The previous version stays active until the issue is resolved.

Zero-Downtime Deployment

Updates are deployed without any service interruption:

1. New version loaded on parallel environment

2. Health check verifies new version

3. Traffic routed to new version

4. Previous version held on standby (rollback)

5. After 15 min stable operation: previous version removed

Total deployment time: under 5 minutes. Customers notice nothing.

Environments

Environment	Address	Purpose	Stability
PROD	`api.muninlabs.io`	Production. Verified releases only	99.9% SLA
DEV	`api.dev.muninlabs.io`	Testing and integration. Identical to prod	Best effort

DEV is identical to PROD The development environment runs the same configuration and software as production. The only difference is that DEV may be updated more frequently and is used for integration testing without risk to production.

Munin Huginn™ — Continuous Improvement

Munin Huginn™ analyses and improves the platform automatically, around the clock:

Task	Frequency	Description
Pronunciation analysis	Every 2 hours	Automatically identifies and corrects pronunciation errors
Training data generation	Daily	Generates new training examples for all supported languages
Model optimisation	Weekly	Evaluates and improves model quality
Quality reporting	Daily	Automatic report on improvements and deviations

API Reference

Munin Labs offers a simple, stable API based on WebSocket (AudioHook protocol) for real-time speech-to-text and text-to-speech.

Endpoints

Production — TTS https://api.muninlabs.io/v1/speak

Production — AudioHook (STT) wss://api.muninlabs.io/v1/ws

Development / Testing https://api.dev.muninlabs.io/v1/speak

Language-Specific Endpoints

For the simplest possible integration we offer language-specific endpoints that enforce the correct language without any language detection. Each endpoint supports both TTS (HTTPS) and AudioHook STT (WSS).

Language	TTS (HTTPS)	AudioHook STT (WSS)	ISO
Norwegian Bokmål	`nb.api.muninlabs.io/v1/speak`	`nb.api.muninlabs.io/v1/ws`	nb
Norwegian Nynorsk	`nn.api.muninlabs.io/v1/speak`	`nn.api.muninlabs.io/v1/ws`	nn
Swedish	`sv.api.muninlabs.io/v1/speak`	`sv.api.muninlabs.io/v1/ws`	sv
Danish	`da.api.muninlabs.io/v1/speak`	`da.api.muninlabs.io/v1/ws`	da
Finnish	`fi.api.muninlabs.io/v1/speak`	`fi.api.muninlabs.io/v1/ws`	fi
Icelandic	`is.api.muninlabs.io/v1/speak`	`is.api.muninlabs.io/v1/ws`	is
Northern Sami	`sme.api.muninlabs.io/v1/speak`	`sme.api.muninlabs.io/v1/ws`	sme
English	`en.api.muninlabs.io/v1/speak`	`en.api.muninlabs.io/v1/ws`	en
German	`de.api.muninlabs.io/v1/speak`	`de.api.muninlabs.io/v1/ws`	de
+ 9 more languages	fr, es, it, nl, pl, hu, pt, ja, ar — same pattern

Customer-Isolated Endpoints

Enterprise customers receive their own fully isolated endpoints. These are provisioned automatically during onboarding:

Format: {language}.{customer}.api.muninlabs.io


            nb.{customer}.api.muninlabs.io — Norwegian Bokmål

            sv.{customer}.api.muninlabs.io — Swedish

            da.{customer}.api.muninlabs.io — Danish

Provisioned automatically during onboarding. Full data isolation, dedicated Munin Accelerate™, and individual SLA monitoring.

Authentication

All API calls require authentication via API key. The key is provided during onboarding.

          # Authentication via Bearer token

          Authorization: Bearer mk_your_api_key_here

Genesys TTS Connector (BYOT-A)

For Genesys Cloud integration the TTS Connector is used. Two endpoints are required:

1. List voices

GET/v1/tts/voicesVoice catalogue for Architect

$ curl https://api.muninlabs.io/v1/tts/voices \

    -H "Authorization: Bearer mk_..."

# Response:

{

  "voices": [

    {"name": "Nora", "id": "nora", "language": "nb-NO", "gender": "female", "type": "neural"},

    {"name": "Erik", "id": "erik", "language": "nb-NO", "gender": "male", "type": "neural"},

    {"name": "Astrid", "id": "astrid", "language": "sv-SE", "gender": "female", "type": "neural"},

    ...

  ]

}

2. Synthesise speech

POST/v1/tts/synthesizeText to PCM audio

$ curl -X POST https://api.muninlabs.io/v1/tts/synthesize \

    -H "Authorization: Bearer mk_..." \

    -H "Content-Type: application/json" \

    -d '{"text":"Welcome to customer service","voice":"nora","language":"nb-NO","sample_rate":8000,"format":"pcm"}' \

    --output audio.pcm

# Response: raw PCM audio (16-bit, mono, 8kHz)

# Content-Type: audio/L16;rate=8000

Genesys Configuration Complete Advanced JSON for Genesys Admin is provided during onboarding. Paste into Admin > Integrations > TTS Connector > Advanced. Credentials: "User Defined" with Authorization header.

REST API (Direct Integration)

For customers who do not use Genesys, we offer a simpler REST API:

Text to Speech

GET/v1/speakGenerates WAV audio

$ curl "https://api.muninlabs.io/v1/speak?text=Hello&voice=nora&language=norwegian" \

    -H "Authorization: Bearer mk_..." \

    --output hello.wav

Parameter	Type	Required	Description
`text`	string	Yes	Text to be spoken (max 2,000 characters)
`voice`	string	Yes	Voice ID (e.g. "nora", "erik")
`language`	string	Yes	Language ("norwegian", "swedish", "danish", etc.)
`speed`	float	No	Speech rate 0.8–1.2 (default: 1.0)

Health Check

GET/v1/healthNo authentication required

$ curl https://api.muninlabs.io/v1/health

{"status":"ok","engine":"Munin Voice Engine","version":"5.0","voices":68}

Voice List

GET/v1/voicesAll available voices

Error Codes

Code	Meaning	Action
`200`	OK — audio returned	—
`400`	Invalid request (missing text/voice)	Check parameters
`401`	Invalid or missing API key	Check Authorization header
`404`	Unknown voice or language	Check /v1/tts/voices for valid values
`429`	Too many requests	Wait and retry (rate limit)
`500`	Internal error	Contact support
`503`	Service temporarily unavailable	Automatic failover active, retry

Language Codes

Language	REST API	Genesys (BCP-47)
Norwegian Bokmål	`norwegian`	`nb-NO`
Norwegian Nynorsk	`norwegian`	`nn-NO`
Swedish	`swedish`	`sv-SE`
Danish	`danish`	`da-DK`
Finnish	`finnish`	`fi-FI`
Icelandic	`icelandic`	`is-IS`
Northern Sami	`northern sami`	`se-NO`
English	`english`	`en-GB`
German	`german`	`de-DE`

Rate Limits

Limit	Value
Max text length per request	2,000 characters
Max requests per minute	Depends on agreement
Max concurrent requests	Depends on agreement

Dashboard

Platform Improvements

Recent Updates

Key Metrics

Platform Overview

Production Architecture

Language Endpoints

Customer-Isolated Endpoints

Security Overview

Encryption

Access Control

Data Residency

Munin Sentinel™ — Continuous Monitoring

Security Policies

GDPR & Privacy

Data Flow Diagram

Privacy Guarantees

Compliance Roadmap

Customer Data Deletion (GDPR Art. 17)

SLA — Service Level Agreement

Failover Architecture

SLA Tiers

Incident Response

SLA Breach Compensation

Development & Deployment Process

Deployment Pipeline

Munin Deploy Gate™

Zero-Downtime Deployment

Environments

Munin Huginn™ — Continuous Improvement

API Reference

Endpoints

Language-Specific Endpoints

Customer-Isolated Endpoints

Authentication

Genesys TTS Connector (BYOT-A)

1. List voices

2. Synthesise speech

REST API (Direct Integration)

Text to Speech

Health Check

Voice List

Error Codes

Language Codes

Rate Limits