Downloads · 30 days
14
3% of all-time downloads
sw_whisper_nonstandard_medium
Veronica1NW/sw_whisper_nonstandard_medium
sw_whisper_nonstandard_medium is a automatic speech recognition model from Veronica1NW. Use it when you need speech turned into text. The card lists the license as apache-2.0.
Languages: Swahili (sw) — Non-standard / accented varieties from Kenyan speakers
All-time downloads
479
Public
Parameters
764M
3.1 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors3.1 GB · 100%
At a glance
- Task
- Automatic Speech Recognition
- License
- apache-2.0
- Model type
- whisper
- Access
- Public
- Created
- Oct 27, 2025
- Updated
- Oct 27, 2025
- SHA
- 7f3d1b0f
Base models
Model card
From the Hugging Face model README
Swahili Whisper Non-Standard Medium
Languages: Swahili (sw) — Non-standard / accented varieties from Kenyan speakers
This model was fine-tuned on non-standard and accented Swahili speech using Common Voice and additional curated speech data from Kenyan speakers.
The goal was to improve ASR performance on regional Swahili accents and non-standard pronunciations.
Fine-tuning data: Subset of Mozilla Common Voice + curated non-standard Swahili speech
📊 Evaluation
| Metric | Pre-train | Test |
|---|---|---|
| WER | 0.2902 | 0.2648 |
| CER | 0.1234 | 0.1069 |
- Task
- Automatic Speech Recognition
- Type
- whisper
- License
- apache-2.0
- Languages
- sw
- Created
- Oct 27, 2025
- Updated
- Oct 27, 2025