KevKibe/Finetuning-WhisperSmall-LoRA-Swahili

This project is a standard implementation of fine-tuning a Whisper model using Hugging Face's PEFT (Parameter-Efficient Fine-Tuning) library. With only 5 stars and 0 forks over nearly 900 days, it lacks community traction and serves primarily as an educational reference or a personal experiment. From a competitive standpoint, it has no defensibility; the approach follows well-documented public tutorials. Frontier labs like OpenAI and Meta have already released much larger and more capable multilingual models (Whisper v3, MMS) that likely outperform a fine-tuned Whisper-Small in Swahili without additional training. Furthermore, specialized ASR providers like Deepgram or AssemblyAI offer Swahili support out-of-the-box, making this specific fine-tuning script obsolete for most production use cases. The displacement risk is high because the base model used (Whisper-Small) has been superseded by much stronger foundation models.