Speechmatics vs ElevenLabs Scribe: Which Speech-to-Text API Delivers?

Speechmatics provides full real-time speaker diarization, with channel diarization also available. ElevenLabs Scribe offers diarization in batch only (its Scribe v2 model) — not in real-time. For live call analytics or voice agents, Speechmatics is the clear choice.

Speechmatics ranks ahead on overall transcription accuracy in head-to-head customer evaluations, including names and numerical data. Speechmatics is also rated stronger on non-English languages such as Arabic, Turkish, and Hebrew, where ElevenLabs Scribe underperforms.

Speechmatics offers true on-premises containers that run on CPU or GPU and can be deployed in secure, air-gapped networks. ElevenLabs Scribe does not currently provide on-premises deployment — it is a cloud-only service.

Phonetic custom dictionaries are a core, deeply integrated feature in Speechmatics. You can define how brand names, product names, and specialist terms are transcribed using phonetics. For ElevenLabs Scribe, custom vocabulary support is a newer, secondary capability — not a core design priority.

Speechmatics has purpose-built medical uplift models across English, French, German, Spanish, and Arabic-English bilingual. These are production-ready for clinical documentation and healthcare workflows — see the Humetrix case study for a proof point. ElevenLabs Scribe has no equivalent dedicated medical-domain models.

Speech-to-text is Speechmatics' primary focus, and has been for well over a decade. ElevenLabs is a text-to-speech company that recently expanded into speech-to-text. If STT accuracy, reliability, and depth matter to your product roadmap, Speechmatics is the specialist partner built for it.

Speechmatics starts from $0.129/hr for Melia batch transcription — all-inclusive, with no separate charges for speaker diarization or custom vocabulary. ElevenLabs Scribe is approximately $0.22/hr for batch and $0.28–0.48/hr for real-time, with pricing that varies by feature usage.

Melia is Speechmatics’ new multilingual speech-to-text model with native code-switching across all 55+ supported languages in a single pass — no per-language model selection needed. It outperforms Deepgram, Microsoft, and AssemblyAI on most FLEURS language benchmarks, making it the strongest option for multilingual content, accented speakers, and global deployments. Priced from $0.129/hr for batch (10 hrs/month free), it’s also the most affordable model in the Speechmatics range. ElevenLabs Scribe is a single general model with no multilingual code-switching support.