Speechmatics

Speechmatics is a speech API provider for turning recorded files and live audio into text, voice-agent interactions and synthesized speech. It suits developers and organizations building multilingual or specialized speech workflows, especially those that need deployment choices beyond a standard cloud service. Its broad language support and enterprise options are compelling, but usage-based prices and plan-dependent deployment make fit and cost worth checking before committing.

Overview

Founded in 2006 and headquartered in Cambridge, United Kingdom, Speechmatics focuses on business use of speech AI rather than a single-purpose transcription app. Its APIs cover speech-to-text, voice agents and text-to-speech, working with both recordings and live audio. That breadth makes it relevant to teams building products or workflows, while the API-centered approach is less suited to someone who simply wants a personal transcription workspace.

Speech-to-text supports more than 55 languages, including mid-sentence language switching. Speechmatics says its real-time service can return results in less than one second. Speaker diarization and custom vocabulary help distinguish who said what and handle specialized terminology, useful in conversations where generic transcripts can miss important distinctions.

Key features

Speechmatics pairs transcription with tools for voice-agent and telephony systems. It names LiveKit, Vapi and Pipecat integrations for voice agents, and Vonage and Jambonz for telephony and contact centers. These connections make it a plausible fit for teams assembling conversational products or contact-center workflows, rather than just processing occasional audio files.

Plan choices reflect different transcription priorities. Melia 1 emphasizes multilingual transcription and code-switching; Oak 1 Medical targets clinical language, including drug names, dosages, abbreviations and procedure terms. Enhanced prioritizes accuracy, custom dictionaries and advanced punctuation, while Standard trades accuracy for a lower per-hour rate. These distinctions are practical: choosing a cheaper tier makes most sense when volume and budget outweigh the need for the highest stated accuracy.

Security commitments include ISO/IEC 27001:2022 accreditation, GDPR and HIPAA compliance, and SOC 2 Type II certification. Data is encrypted at rest with AES 256 and in transit using TLS 1.2 or higher. Model-training data logging is off by default for self-serve and enterprise customers, an important point for organizations handling sensitive audio.

Pricing

Speechmatics is freemium, with usage-based charges for most paid transcription and synthesis. The Free plan costs nothing and provides $100 in credits without requiring a credit card. It allows 2 concurrent real-time sessions and 10 pre-recorded files per second, with community, portal chat and email support. The credit allocation makes it a sensible way to evaluate workflows, but the low real-time concurrency limits its usefulness for a live service with several simultaneous sessions.

• Text to Speech: $0.01, paid once, billed Usage-based, with the first 1 million characters free. It provides natural, low-latency speech synthesis for voice AI use cases; the free character allowance offers room to assess synthesis before usage charges apply.
• Melia 1: $0.12, paid once, billed Usage-based per hour of pre-recorded transcription. It includes 56 languages, mixed-language transcription, code-switching and speaker diarization, making it the clearest fit for multilingual recordings.
• Oak 1 Medical: $0.15, paid once, billed Usage-based per hour of pre-recorded transcription. Clinical language support covers drug names, dosages, abbreviations and procedure terms, with multilingual support; the specialty is aimed at medical audio rather than general-purpose savings.
• Standard: $0.23, paid once, billed Usage-based per hour of pre-recorded transcription. It has the same features as Enhanced at a lower rate but lower accuracy, making it the cost-sensitive option for high-volume work.
• Enhanced: $0.38, paid once, billed Usage-based per hour of pre-recorded transcription. It offers the highest accuracy, language selection per transcription, a custom dictionary and advanced punctuation for work where transcript quality matters more than the lower Standard rate.
• Enterprise: custom pricing, with unlimited scale, a dedicated customer success manager and solutions engineer, and flexible deployment options. It is intended for organizations needing tailored scale or deployment rather than a self-serve rate.

Self-serve accounts can run up to 50 concurrent real-time sessions, except Free accounts, which are capped at 2; enterprise concurrency adjusts dynamically to demand. There is no free trial, so the free credits are the available starting point for testing. Deployment choices include SaaS, multi-region cloud, on-premises containers and virtual appliances, on-device, and hybrid cloud plus on-premises options, with availability varying by plan. Support expands from community, portal chat and email for self-serve to a dedicated customer success manager and solutions engineer for enterprise.

Platforms

Speechmatics supports API, Linux, macOS, self-hosted, web and Windows. The platform range and deployment options suit teams integrating speech into software or operating it across different environments. Because deployment availability varies by plan, organizations with a requirement for on-device, on-premises or hybrid operation should confirm the relevant plan before choosing on price alone.

Who it's for

Speechmatics names healthcare, voice agents, live captioning, contact-center analytics, legal transcription and meeting platforms among its use cases. Its multilingual handling and speaker labeling are particularly relevant when recordings contain multiple speakers or language changes. The medical tier serves clinical terminology needs; the broader tiers suit teams balancing accuracy, language coverage and processing cost. It is less compelling for individuals seeking a ready-made note-taking app, since its core offering is APIs and usage-priced services.

Pros and cons

• Pro: More than 55 supported languages and code-switching support make it useful for multilingual audio that a single-language workflow would struggle to accommodate.
• Pro: Recorded transcription, real-time speech-to-text, voice agents and text-to-speech bring several speech workflows under one provider.
• Pro: Medical and accuracy-oriented tiers let buyers match spend to specialized vocabulary or transcript quality needs.
• Pro: Multiple cloud, self-hosted, on-device and hybrid deployment models, alongside substantial security commitments, suit organizations with operational or data-protection requirements.
• Con: Most paid options are usage-based, so ongoing costs depend on transcription hours or synthesis volume rather than a simple fixed subscription.
• Con: Free real-time use is limited to 2 concurrent sessions, which is restrictive for testing or serving several simultaneous live streams.
• Con: Deployment availability varies by plan, so buyers cannot assume every deployment model comes with every tier.

Alternatives

For a wider comparison, browse Speech-to-Text Software, Speech Recognition Software, Audio Transcription Software and Podcast Transcription Software.

• Google Cloud Speech-to-Text is worth considering when its paid, API-oriented service and free plan suit the intended recognition workload.
• AssemblyAI offers a free-credits option with limits of 5 new streaming connections per minute and 5 concurrent pre-recorded transcriptions; consider it when those allowances suit an initial evaluation.
• Notta is a freemium option with Android, iOS, desktop and web platforms; its free tier allows 1 seat and 120 transcription minutes per month, a more app-oriented choice for an individual user.
• Amazon Transcribe has a free tier of 60 minutes per month for 12 months, with unused monthly usage not rolling over; it may suit a reader whose needs fit that allowance.
• Happy Scribe offers a free plan with unlimited meeting recordings, 45 minutes per recording and a 10-minute AI trial; consider it for meeting recordings within those limits.
• Rev AI is an API-only alternative with a free plan, suitable for readers seeking another API provider.
• MacWhisper offers a free macOS/iOS option with transcription and subtitle export, plus a €64 a month plan billed Pay Once; it is a better fit for Apple-device users wanting a transcription app.
• Transkriptor offers a Lite plan at $9.99 a month, billed monthly, with 300 minutes of transcription; consider it when a fixed monthly allowance suits individual use.

Verdict

Speechmatics is a strong choice for developers and businesses building multilingual speech products or processing specialized audio, especially when they value speaker identification, medical vocabulary or deployment flexibility. Its breadth of APIs and security posture are the main reasons to choose it. Look elsewhere if you want a straightforward personal transcription app, or if usage-based billing, the Free plan’s two-session cap or plan-dependent deployment options do not fit your needs.