# Speechmatics Speechmatics provides speech-to-text APIs for pre-recorded and real-time audio, with support for multilingual speech, multiple speakers, and voice-agent workflows. It is designed for teams building transcription and Voice AI applications. - Findings24: - Site: - Preços: Freemium - Categorias: Voz para Texto, API de áudio com IA ## O que é Speechmatics? Speechmatics is a speech-to-text API platform for transcribing live or pre-recorded audio and video. Its models are designed for conversational audio, including accents, multiple speakers, multilingual speech, and language switching within a sentence. The platform offers general-purpose, enhanced, medical, and voice-agent-oriented models for different transcription requirements. Core capabilities include batch transcription from audio or video files and real-time transcription from live audio or video. Depending on the model, outputs and processing options include speaker diarization, language hints and labeling, smart formatting, word-level timings, custom dictionaries, healthcare vocabulary handling, mixed-language transcription, and code-switching support. ## O que Speechmatics pode fazer? ### Pre-recorded and real-time transcription Transcribe audio or video files in batch workflows, or process live audio and video for applications that need streaming speech recognition. ### Multilingual and mixed-language speech The Melia 1 model supports 56 languages without requiring a language choice up front and can switch languages mid-sentence. The platform site describes support for more than 55 languages overall. ### Speaker and language labeling Speaker diarization separates contributions in multi-speaker recordings, while language hints and labeling help identify or guide the languages present in a transcription. ### Model options for different domains Available model choices include Melia 1 for multilingual audio, Enhanced for higher-accuracy single-language transcription, Oak 1 for clinical language, and Linden 1 for real-time voice-agent use cases. ### Conversation and document processing add-ons Optional processing includes translation, chapters, topics, summaries, sentiment, PII redaction, and audio alignment. Availability and rates vary by model and processing mode. ### Voice-agent turn detection The Linden 1 real-time model includes turn detection for conversational voice-agent workflows, helping an application identify when a speaker's turn has changed. ## Casos de uso ### Multilingual conversation transcription Use Melia 1 to transcribe meetings, interviews, calls, or other recordings where speakers use multiple languages or switch languages during a conversation. ### Clinical dictation and documentation Use the Oak 1 medical model for clinical language, including drug names, dosages, abbreviations, and procedure terms, with multilingual support. ### Real-time voice agents Use the Linden 1 real-time model as the speech-recognition layer for conversational agents that need built-in turn detection. ### Recorded media processing Process audio or video files in batch and add outputs such as chapters, topics, summaries, translation, or sentiment analysis where the selected model and workflow support them. ## Perguntas frequentes ### Can Speechmatics transcribe live audio as well as recorded files? Yes. The pricing information distinguishes pre-recorded transcription from real-time transcription. Pre-recorded workflows accept audio or video files, while real-time workflows process live audio or video. ### How many languages does Speechmatics support? The site describes support for more than 55 languages. The Melia 1 model is listed as transcribing 56 languages without choosing one language up front and can switch languages mid-sentence. ### Does Speechmatics identify different speakers? Speaker diarization is listed as an available transcription feature. Exact feature availability can differ by model, product package, and deployment. ### Which model is intended for medical transcription? Oak 1 is the medical model. It is intended for clinical language, including drug names, dosages, abbreviations, and procedure terms. ### Is there a free way to start? The pricing page says users can start free with $100 in credits and no card required. Usage rates after the included credits vary by model and processing mode. ## Informações rápidas - **Product type**: Speech-to-text APIs - **Input workflows**: Pre-recorded audio or video and real-time audio or video - **Language coverage**: 55+ languages; Melia 1 lists 56 languages - **Notable models**: Melia 1, Enhanced, Oak 1, and Linden 1 - **Deployment**: Cloud SaaS; the product site also describes cloud, on-premises, and on-device options - **Pricing model**: Usage-based pricing with a free start offer and model-specific rates