Speechmatics Pricing 2026

Speechmatics offers a Free tier with 480 minutes/month of speech-to-text (recurring monthly) and 1 million characters/month of text-to-speech (English only), requiring no credit card. The Pro tier costs $0.24/hour for standard speech-to-text and includes the same 480 free minutes/month. Text-to-speech on Pro is English-only; multilingual TTS requires Enterprise. Enterprise pricing is custom-quoted and includes custom models, on-premises deployment, unlimited scale, and dedicated support.

Yes, Speechmatics offers a Free tier with 480 minutes/month of speech-to-text and 1 million characters/month of text-to-speech (English only), with no credit card required. This free tier resets monthly, unlike competitors' one-time credits (Deepgram's $200, AssemblyAI's $50). At 480 minutes/month, the Free tier provides 5,760 minutes/year (96 hours) of ongoing transcription. For teams processing more than 8 hours/month, the Pro tier at $0.24/hour is the next step. For larger free testing budgets, Deepgram's $200 credit (~433 hours one-time) is more generous upfront.

Speechmatics is a speech recognition platform providing speech-to-text and text-to-speech APIs for developers. It specializes in multilingual transcription with 55+ languages, on-premises deployment options for data sovereignty, and per-second billing precision. Speechmatics differentiates itself through its recurring free tier (480 minutes/month, unlike competitors' one-time credits), automatic volume discounts (20% above 500 hours/month), and enterprise-grade deployment flexibility (cloud, on-premises, or air-gapped). The platform is used by companies like BBC, Sky, and Vodafone for live captioning, call analytics, and media transcription.

Speechmatics Pro at $0.004/min ($0.24/hr) is 60% more expensive than AssemblyAI Universal at $0.0025/min ($0.15/hr) for basic transcription. However, Speechmatics includes 480 minutes/month free that resets monthly (5,760 min/year ongoing), while AssemblyAI's $50 credit is one-time (~185 hours). Speechmatics supports 55+ languages natively, while AssemblyAI focuses on English and common languages. AssemblyAI offers richer audio intelligence features (entity detection, topic detection, auto chapters), while Speechmatics excels at multilingual transcription and on-premises deployment. Choose Speechmatics for ongoing free tier, multilingual support, or on-premises needs; choose AssemblyAI for audio intelligence features and lower base pricing.

Speechmatics Pro at $0.24/hour is about 48% cheaper than Deepgram Nova-3 at $0.0077/min for batch transcription. Speechmatics includes 480 minutes/month free that resets monthly (5,760 min/year), while Deepgram offers $200 in one-time credits (~433 hours). Deepgram excels at real-time streaming with <300ms latency, which Speechmatics also supports. Speechmatics supports 55+ languages vs Deepgram's 30+. Speechmatics offers on-premises deployment (Enterprise), while Deepgram is cloud-only. Choose Speechmatics for lower per-hour cost, multilingual support, recurring free tier, or on-premises deployment; choose Deepgram for ultra-low latency real-time streaming, larger upfront free credit, or Voice Agent API.

Speechmatics' Free tier includes 480 minutes/month of speech-to-text (55+ languages, batch and real-time), 1 million characters/month of text-to-speech (English only), and 2 concurrent real-time sessions. The Pro tier ($0.24/hr for standard speech-to-text) includes the same 480 free minutes/month plus 50 concurrent real-time sessions, 10 file jobs/second for batch processing, and email support. Pro usage is capped at 6,000 hours/month. Enterprise includes custom models, on-premises deployment, unlimited scale, dedicated support, and custom SLA. All tiers bill per-second with no rounding.

Speechmatics operates on a usage-based billing model for Pro tier with no subscriptions or advance payments, so there are no refunds -- you are billed only for audio processed beyond the 480 free minutes/month. The Free tier includes 480 minutes/month that resets monthly with no charges. Enterprise customers with prepaid annual commitments should negotiate refund terms directly in their contracts, as prepaid credits may expire annually and are typically non-refundable for unused balances. If you encounter a service issue or are overcharged due to a bug, contact Speechmatics support to request a credit adjustment.

Yes, Speechmatics offers a recurring free tier with 480 minutes/month (8 hours/month) of speech-to-text and 1 million characters/month of text-to-speech (English only) that resets monthly. This provides 5,760 minutes/year (96 hours) of ongoing transcription at no cost, making it a strong free tier for sustained low-volume usage. Deepgram's $200 credit (~433 hours) is more generous upfront but does not refresh. AssemblyAI's $50 credit (~185 hours) is one-time. For teams processing under 8 hours/month long-term, the Speechmatics Free tier is the most cost-effective option. For production workloads exceeding 8 hours/month, the Pro tier at $0.24/hour is the next step.

Based on available user feedback, the Free plan has reported stability concerns: usage limits may be applied without warning at levels far below what is advertised, and transcription may silently fail on sections of audio without returning error messages. For production workloads requiring reliability guarantees, Pro or Enterprise — with explicit SLA terms — is recommended.

Beyond the advertised usage limits, Free plan users have reported three non-obvious restrictions: (1) effective usage caps are applied at more than 10x below the advertised limits with no warning, (2) the service may silently downgrade to lower-quality transcription models without notification, and (3) transcription jobs can silently skip large sections of audio without returning any error — producing incomplete output that looks complete.

Upgrade from the Free plan to Pro if you need: predictable usage capacity without undisclosed soft caps, consistent guaranteed model quality, or error transparency and reliability SLAs for any production pipeline. The Free plan is not appropriate for workloads where silent transcription failures would go undetected.