speechServer MRCP
Network-based MRCPv2 server for IVR text-to-speech, certified for Cisco CVP and Avaya AAEP. 150+ AI voices in 50+ languages, on-premise, with ReadSpeaker support included.
Certified interoperability
Certified fully compliant with the two IVR platforms most contact centers run, and standards-based everywhere else.
Cisco CVP (Cisco CVP 12.x · CVVB)
Certified fully compliant
Avaya AAEP (Avaya Experience Portal 8.x · AAEP 7.x/8.x)
Certified fully compliant
Genesys GVP (Genesys GVP 8.x · 9.x)
Used in conjunction with
UniMRCP (UniMRCP · IETF MRCPv2)
Standards-based interoperability
Who it's for
Integrate text-to-speech into Cisco CVP or Avaya AAEP, evaluate it against your RFP, and wire it up over MRCPv2, with standards-based interoperability and ReadSpeaker support throughout.
Drop MRCPv2 text-to-speech into your Cisco CVP or Avaya AAEP stack with no proprietary adapters. UniMRCP interoperability lets any standards-based client connect. One server delivers 150+ AI voices across 50+ languages.
Every specification on this page is documented and ready for your procurement review. ReadSpeaker is an ISO/IEC 27001:2022 certified organization, GDPR-aligned with EU Standard Contractual Clauses. The license covers voices, ports and term, so scope stays fixed and the contract stays predictable, with implementation support included.
Send plain text or SSML in and get A-law or μ-law PCM back. Switch voice and language inline, add IPA dictionaries for tricky pronunciations, insert audio clips, and fire mark events for IVR synchronization. Command-line tools and a Test Client ship with every install.
Use cases
From banking IVR to public service helplines, speechServer MRCP delivers the text-to-speech layer your contact center runs on. It speaks 50+ languages over the standards your platform already uses.
Trust starts with the voice on the line
Dynamic IVR prompts read account details, transaction confirmations and fraud notifications in the caller's language with natural-sounding voices. Multilingual coverage maintains brand consistency across markets.
Runs on
Comprehension is care
Outbound IVR confirms appointments, refills prescriptions and reminds patients of pre-procedure steps. Because it runs on-premise, patient health data stays internal, and 50+ languages reach every patient population.
See healthcare sectorEvery announcement, every language
Voice real-time platform announcements, dynamic timetables, ticketing kiosk prompts and IVR rerouting. Telephony-grade audio formats work with legacy infrastructure, in 50+ languages.
See transportation sectorPublic information that reaches every citizen
Multilingual IVR powers citizen helplines, utility outage updates, post-call satisfaction surveys and regulatory voice notifications. EU data sovereignty stays intact, because input text and generated audio never leave your infrastructure.
Runs on
Features
A standards-based MRCP server your IVR platform plugs into without custom adapters, sized for production telephony, with 50+ languages out of the box.
Native MRCPv2 and UniMRCP support means Cisco CVP and Avaya AAEP are certified out of the box, with Genesys GVP and UniMRCP-based platforms connecting via the same standards-based path over TCP/IP. No proprietary glue code.
Clear, natural, human-like Deep Neural Network voices, continually optimized for consistent quality. New voices and languages added regularly to the catalog.
Multithreaded architecture handles multiple concurrent TTS requests, with concurrency determined by your licensed number of ports and available system resources. Memory-efficient channels (2-90 MB each) keep server costs predictable as you scale.
Handle multilingual callers in a single session. SSML instructions let you switch voices or languages inline, without ending the call or reconnecting.
Build per-language pronunciation dictionaries for product names, acronyms and proper nouns. IPA transcription gives you precise phonetic control when you need it.
Reference pre-recorded audio files inline with your TTS prompts. The server streams everything back in the correct order, useful for hold music, legal notices and brand jingles.
Mark information fires events at specific text positions, so your IVR can advance screens, log interactions or sync animations precisely with the spoken audio.
Adjust speaking rate, pitch and volume per prompt, from routine support messages to urgent alerts. Configurable per request via SSML or in your default server config.
A built-in command-line tool handles start, stop, version checks and server operations, ready for your monitoring, deployment and incident response workflows.
Technical specifications
Everything your operations and IVR team needs to validate the integration, sized for production telephony traffic.
| Specification | Value |
|---|---|
| Protocol | IETF MRCPv2 (draft-ietf-speechsc-mrcpv2-21) over TCP/IP. UniMRCP interoperability. |
| Voices | 150+ voices, 50+ languages. Deep Neural Networks technology. |
| Windows OS | Windows Server 2019 and higher. Most recently tested: Windows Server 2025. |
| Linux OS | CentOS Stream 9.x / 10.x · Ubuntu 18.x to 24.x · RHEL 8.x / 9.x · Rocky Linux 9.x |
| CPU | Intel x64, 1 GHz or higher. If the AVX instruction set is not supported, consult ReadSpeaker. |
| RAM | 4 GB or more recommended (depending on the number of channels). |
| Voice footprint | 9 to 35 MB per voice. |
| Runtime memory | 100 MB basic load · 35 MB per voice · 2 to 90 MB per channel. |
| Input formats | Plain text (multibyte encoding) · SSML (UTF-8 encoding). |
| Audio output formats | 8 KHz 8-bit A-law PCM · 8 KHz 8-bit μ-law PCM (telephony-grade) |
| IVR platform compliance | Certified compliant: Cisco CVP 12.x · Cisco CVVB · Avaya Experience Portal 8.x · Avaya AAEP 7.x / 8.x. Used in conjunction with: Genesys GVP 8.x / 9.x. |
| Configuration settings | SIP port · MRCPv2 port · RTP port · Volume · Speaking rate · Pitch · Sentence pause time · Comma pause time · Default user dictionary. |
| Solution components | ReadSpeaker speechServer MRCP · ReadSpeaker TTS Engine · User Dictionary · User Documentation. |
| Documentation | Product Specification (PDF) |
How it works
speechServer MRCP runs as a standards-based MRCPv2 server on your infrastructure. Integration follows the IETF spec, with implementation support from the ReadSpeaker team end to end.
Run the speechServer MRCP installer on Windows Server 2019+ or Linux (CentOS Stream, Ubuntu, RHEL, Rocky Linux). The installer deploys the TTS engine, voice data files, Test Client and command-line tools.
Apply the license file ReadSpeaker provides. It enforces licensed voices, concurrent TTS port capacity, synthesis speed and license term. Configure SIP, MRCPv2 and RTP ports plus DSP defaults in the config file.
Point Cisco CVP, Avaya AAEP, Genesys GVP or a UniMRCP-based client at the speechServer MRCP endpoint. The client requests TTS synthesis over TCP/IP per IETF MRCPv2 spec, with no proprietary adapters.
The server returns 8 KHz A-law/μ-law PCM audio streams to the IVR. Switch voices and languages inline via SSML. Insert pre-recorded audio clips. Fire mark events for client-side synchronization.
Compliance
ReadSpeaker is an ISO/IEC 27001:2022 certified organization, GDPR compliant, with Standard Contractual Clauses for cross-border data transfers. speechServer MRCP runs on your own infrastructure: input text and generated audio remain in your environment by design, never leaving the perimeter you control.
ISO/IEC 27001:2022 (ISO/IEC 27001:2022)
ReadSpeaker organization
GDPR (GDPR)
Privacy Notice published
On-premise (On-premise)
No data leaves your network
EU based (EU based)
Headquartered in Sweden (ReadSpeaker AB, Uppsala)
FAQ
ReadSpeaker speechServer MRCP is a network-based text-to-speech solution that enables MRCP-compliant client applications to request high-quality speech synthesis within a client-server architecture. It provides standards-based interoperability with MRCPv2 and UniMRCP servers for IVR environments. Available on Windows Server 2019+ and Linux.
speechServer MRCP supports IETF MRCPv2 (draft-ietf-speechsc-mrcpv2-21) over TCP/IP, with interoperability for UniMRCP servers.
Yes. ReadSpeaker speechServer MRCP is certified fully compliant with Cisco CVP 12.x and Cisco Virtualized Voice Browser (CVVB).
Yes. speechServer MRCP is certified fully compliant with Avaya Experience Portal 8.x and Avaya AAEP (Aura Experience Portal) 7.x/8.x.
speechServer MRCP adheres to the standards of and is used in conjunction with Genesys GVP 8.x and 9.x. This configuration is not listed in the formal certified compliance matrix, which covers Cisco CVP, Cisco CVVB and Avaya AAEP. To confirm support scope for Genesys, contact your ReadSpeaker account team.
Yes. speechServer MRCP supports MRCPv2 and UniMRCP interoperability, so any standards-based UniMRCP client can connect over TCP/IP. The certified compliance matrix formally lists Cisco CVP, Cisco CVVB and Avaya AAEP, with Genesys GVP used in conjunction. For any platform outside that matrix, contact your ReadSpeaker account team to confirm support scope.
They solve different parts of the stack. UniMRCP is an open-source MRCP framework that connects to a separate text-to-speech engine of your choice. ReadSpeaker speechServer MRCP is a complete commercial TTS solution: it bundles the ReadSpeaker TTS engine and 150+ Deep Neural Network voices, includes license file enforcement and implementation support from the ReadSpeaker team, and is certified fully compliant with Cisco CVP and Avaya AAEP. speechServer MRCP can itself interoperate with UniMRCP servers over standards-based MRCPv2.
speechServer MRCP is licensed under a license agreement and priced by the number of concurrent TTS ports you need. A license file enforces the licensed voices, port capacity, synthesis speed, number of servers covered and license term. Contact ReadSpeaker for a tailored quote.
speechServer MRCP supports 8 KHz 8-bit A-law PCM and 8 KHz 8-bit μ-law PCM, telephony-grade formats compatible with standard IVR systems.
Windows Server 2019 and higher (most recently tested: Windows Server 2025). Linux: CentOS Stream 9.x to 10.x, Ubuntu 18.x to 24.x, RHEL 8.x to 9.x, Rocky Linux 9.x.