MERaLiON GI v1 / GR
The official GI v1 repository publishes the speech classifier described as MERaLiON-GR. It fine-tunes SpeechEncoder 2 for a two-label classification task. The differing GI and GR names do not establish two separate models.
A task-specific encoder adaptation
The system uses LoRA on the speech encoder, learned aggregation of hidden states, an ECAPA-TDNN network and attention pooling. Its paper evaluates English and several Southeast Asian languages. This trained classifier is distinct from the AudioLLM used as an evaluation baseline.MERaLiON-GR technical report
Use the published repository identifier
The card heading uses GR and an example uses GR-v1, but the available repository is GI-v1. The linked analysis service currently names SER v1, so an exact hosted GI checkpoint mapping remains unconfirmed.MERaLiON GI v1 / GR official model cardMERaLiON speech analysis application
The GI repository includes a 2026 custom license that explicitly lists this model. It retains upstream conditions and adds notice and acknowledgement duties. Its component-license list is not, by itself, an architecture specification.MERaLiON GI v1 repository license, 2026
Keep the label definition visible
Practical guidance. The output is a prediction within a binary training scheme. It cannot establish a person's self-identified gender or biological sex. If studying acoustic classification, state the dataset's label method and assess errors across recording conditions. Do not present the labels as verified personal facts.MERaLiON-GR technical report
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | MERaLiON-GI-v1MERaLiON-GRMERaLiON-GR-v1 |
| Published labels | Female and male, represented as two output classes.MERaLiON GI v1 / GR official model card |
| Input | Speech audio sampled at 16 kHz.MERaLiON GI v1 / GR official model card |
| Encoder ancestry | MERaLiON-SpeechEncoder-2.MERaLiON GI v1 / GR published configuration |
Related models
- MERaLiON SER v1
MERaLiON / A*STAR IAIC · MERaLiON Speech Emotion Recognition
A 309M speech classifier that uses Whisper-Medium features to estimate emotions and continuous affect values.
Launch date not confirmedDownloadable · API · Provider product - MERaLiON SpeechEncoder 2
MERaLiON / A*STAR IAIC · MERaLiON SpeechEncoder
A 630M-class multilingual speech encoder trained from scratch to provide acoustic features for downstream tasks.
Launch date not confirmedDownloadable