---
language:
- ms
- en
- zh
- ta
datasets:
- mesolitica/Malaysian-STT-Whisper
- malaysia-ai/STT-Whisper
base_model:
- openai/whisper-large-v3-turbo
---

# Malaysian Finetune Whisper Large V3 Turbo

Finetune Whisper Large V3 Turbo on Malaysian context.

## Improvement

1. Distilled from Whisper Large V3 on Malaysian and Science context.
2. Better translation for Malay, Manglish, Mandarin, Tamil and Science context.
3. Word level timestamp, introduced `<|transcribeprecise|>` token, **a new task!**

## Example

We use [assembly.mp3](https://github.com/mesolitica/malaya-speech/blob/master/speech/assembly.mp3) and [toodia.mp3](https://github.com/mesolitica/malaya-speech/blob/master/speech/podcast/toodia.mp3) for examples.

### Whisper Large v3

### Whisper Large v3 Turbo

### mesolitica/Malaysian-whisper-large-v3-turbo-v3

## how we finetuned it?

We done 2 phases,

1. Finetune on [mesolitica/Malaysian-STT-Whisper](https://huggingface.co/datasets/mesolitica/Malaysian-STT-Whisper)
- Revision [267552e0f093068519a816112c2741939d057f48](https://huggingface.co/mesolitica/malaysian-whisper-large-v3-turbo-v3/commit/267552e0f093068519a816112c2741939d057f48)
- WanDB at https://wandb.ai/huseinzol05/malaysian-whisper-large-v3-turbo-v3
2. Stage 2, Annealing on 5% from [mesolitica/Malaysian-STT-Whisper](https://huggingface.co/datasets/mesolitica/Malaysian-STT-Whisper) and 100% from [mesolitica/Malaysian-STT-Whisper-Stage2](https://huggingface.co/datasets/mesolitica/Malaysian-STT-Whisper-Stage2)
- Revision [5f6ca0596f01527ec2d013662cfc168d5f754461](https://huggingface.co/mesolitica/malaysian-whisper-large-v3-turbo-v3/commit/5f6ca0596f01527ec2d013662cfc168d5f754461)
- WanDB at https://wandb.ai/huseinzol05/malaysian-whisper-large-v3-turbo-v3-stage2

## Benchmark