Speechmatics vs Google Cloud Speech-to-Text: Which API Delivers?

Speechmatics scores 92% accuracy (G2 Spring 2026) compared to Google’s 87%. Speechmatics is trained on over a million hours of noisy, accented, real-world audio.

Yes — on-premises, on-device, and fully air-gapped with no cloud dependency. Google's Chirp 3 runs only in Google Cloud; Google's on-premises and air-gapped speech options (Speech-to-Text On-Prem and Google Distributed Cloud) are sales-led and use older models, not Chirp 3.

No. Speechmatics is cloud-agnostic. Google Cloud Speech-to-Text ties your pipeline to Google Cloud.

Speechmatics starts from $0.12/hr (~$0.002/min) for Melia batch — simple, all-inclusive per-hour pricing. Google Cloud Speech-to-Text (V2, including Chirp 3) lists $0.016/min ($0.96/hr) for both streaming and standard batch for the first 500,000 minutes a month, with volume tiers down to $0.004/min ($0.24/hr) above 2 million minutes. Its dynamic batch option is $0.003/min ($0.18/hr), still above Melia's batch rate.

Yes — low-latency streaming with diarisation included at no extra charge, and no session time limit. Google Cloud Speech-to-Text streaming sessions timeout after approximately 5 minutes, requiring reconnection for long-running audio. Speechmatics’ real-time capability is built for production voice agent deployments where uninterrupted, speaker-aware streams are essential.

Speechmatics supports 55+ production-proven languages with one inclusive model per language — all regional variants included, no manual dialect sub-model selection required. Google lists 85+ languages and variants (most Chirp 3 locales are still in Preview) and transcribes per locale, such as es-ES vs es-US; Chirp 3's automatic language detection transcribes the most prevalent language in the audio.

Yes — ISO 27001, SOC 2, HIPAA, and GDPR, with dedicated enterprise support and SLAs.

Yes. Speechmatics includes a dedicated Medical Model purpose-built for clinical documentation — trained on SNOMED CT terminology, FDA/MHRA drug names, and real clinical audio. It delivers up to 50% fewer critical errors compared to general models, with 96% medical term recall, available for real-time transcription across English, French, German, Spanish, Danish, and Norwegian. Google’s medical models (medical_conversation and medical_dictation) are available in US English only. See the Medical Model launch announcement and the Humetrix case study — where Speechmatics replaced Whisper for multilingual clinical transcription across 27 languages at the Paris 2024 Olympics.

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, which makes it a strong fit for multilingual content, accented speakers, and global deployments. Priced from $0.12/hr for batch, it’s also the most affordable model in the Speechmatics range, and new accounts get $100 in credit to get started. Google’s Chirp 3 transcribes the most prevalent language in the audio, whereas Melia handles code-switching natively.

Chirp 3 is the latest speech-to-text model in Google Cloud Speech-to-Text, succeeding Chirp 2. It's available only through the Speech-to-Text V2 API on Google Cloud, and its speaker diarization works in batch only, not streaming. Speechmatics' Ursa 2 covers real-time and batch with one model per language and all regional accents included, and Melia adds multilingual batch transcription with native code-switching across 55+ languages. Speechmatics also runs on-premises and air-gapped.

Gemini can transcribe audio and, when prompted, label speakers and add segment-level timestamps, but it's a general multimodal model rather than a dedicated speech-to-text API. Google itself points users to Cloud Speech-to-Text for dedicated real-time transcription. Speechmatics is purpose-built for speech-to-text, with real-time diarization, custom dictionaries and on-premises deployment.