MERaLiON SER v1
MERaLiON SER v1 is a 309M-parameter speech classification model. It adapts Whisper-Medium acoustic features to estimate emotion categories and continuous affect values. It is a trained classifier, not a transcription model.
A trained speech classifier
The model uses LoRA adapters on Whisper-Medium's encoder attention projections. A modified ECAPA-TDNN and attention pooling feed two prediction heads. The Whisper decoder is unused. This architecture estimates acoustic attributes without generating a transcript.MERaLiON SER v1 official model card
Weight download and hosted analysis
The analysis application explicitly names SER v1. The OpenAPI specification names MERaLiON-SER for its combined emotion-and-gender endpoint. This establishes a hosted route, but does not prove that every service output comes from one unchanged weight file.MERaLiON speech analysis applicationMERaLiON public OpenAPI specification
The SER-specific license contains MIT terms plus additional redistribution and rights conditions. The separate API disclaimer governs hosted use.MERaLiON SER v1 Public Licence
Treat outputs as estimates
Practical guidance. Review errors by language and emotion class. A high overall score can hide weak performance on rare labels. Use the output as an uncertain acoustic estimate, not proof of a person's feelings. The API terms reject using such estimates as the main basis for consequential decisions.MERaLiON API terms of use
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable, API, Provider product |
| Model identifiers | MERaLiON-SER-v1MERaLiON-SER |
| Outputs | Seven emotion classes plus valence, arousal and dominance values.MERaLiON SER v1 official model card |
| Primary language coverage | English, Chinese, Malay and Tamil; limited Thai, Indonesian and Vietnamese support.MERaLiON SER v1 official model card |
| Hosted clip limit | The public analysis application accepts up to five minutes.MERaLiON speech analysis application |
Related models
- MERaLiON 3 10B
MERaLiON / A*STAR IAIC · MERaLiON 3
A third-generation speech-text model with regional fine-tuning for audio questions, dialogue summaries and contextual analysis.
Launch date not confirmedDownloadable - MERaLiON GI v1 / GR
MERaLiON / A*STAR IAIC · MERaLiON Speech Gender Recognition
A two-label speech classifier derived from SpeechEncoder 2, published as GI v1 and described as MERaLiON-GR.
Launch date not confirmedDownloadable