Melia 1
Melia 1 produces one transcript from speech that changes language. Its June 2026 launch covered more than 55 languages and started with a Batch production preview.
Language changes within the same speech
The Realtime release labels each transcribed word with its language. Optional language hints bias recognition toward expected languages. A LanguageInfo event announces each newly detected language during the session.Speechmatics release notes: Melia 1 Realtime, 13 October 2026
The September evaluation examined mixed error rate and switch-point F1 separately. This distinguishes getting words right from detecting where a speaker changes language. The same report also described improvements to spoken reference numbers and other alphanumeric strings.Speechmatics: Melia code-switching evaluation
Practical guidance. A mixed-language transcript needs both correct words and correct language boundaries. These are useful separate checks when captions, routing or later text processing depend on the language labels.Speechmatics: Melia code-switching evaluation
Current access and limits
The container guide describes Melia as GPU-only. Batch is generally available there, while container streaming remains an early preview. A container preview label does not describe every hosted API route.Speechmatics documentation: Accessing container images
The launch positioned Melia alongside Standard and Enhanced. It did not replace those models or promise the same feature set.Speechmatics: Introducing Melia
Melia lacks custom dictionary, confidence scores and speaker identification. Batch supports speaker diarization; Realtime currently supports channel diarization only.Speechmatics documentation: Models
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API, Provider product |
| Model identifiers | melia-1 |
| API configuration | model: melia-1; language: multiSpeechmatics release notes: Melia 1 Realtime, 13 October 2026 |
| Availability | Batch generally available; Realtime SaaS released 13 October 2026Speechmatics documentation: Accessing container imagesSpeechmatics release notes: Melia 1 Realtime, 13 October 2026 |
Related models
- Oak 1
Speechmatics · Oak
Multilingual medical recognition model for consultations that change language.
Launch date API · Provider product - Mandarin–English Bilingual
Speechmatics · Bilingual recognition
Bilingual recognition model trained for Mandarin and English code-switching in Southeast Asia.
Launch date API · Provider product - Enhanced
Speechmatics · Enhanced
General speech recognition model that prioritizes accuracy across supported language packs.
Launch date not confirmedAPI · Provider product - Standard
Speechmatics · Standard
General speech recognition model that prioritizes transcription throughput.
Launch date not confirmedAPI · Provider product