Spark-TTS Acholi Fine-tune -- v11
This model is part of a 16-experiment series fine-tuning Spark-TTS for Acholi (ach) text-to-speech, conducted as part of ongoing Masters research on offline AI dubbing for low-resource African languages.
Model weights and detailed experimental findings will be published after thesis submission.
Full Documentation
For the complete experiment log, methodology, and evaluation results across all 16 experiments, see the canonical model:
acellam/spark-tts-salt-ach-v16
About This Experiment
Version: v11 -- Fixed data pipeline
Language: Acholi (ach) -- Nilotic (Luo branch), ~1.5M speakers, northern Uganda and South Sudan
Pipeline: Whisper ASR -> SALT NLLB-1.3B Translation -> Spark-TTS Synthesis
Training scripts: gitlab.com/mistaguy6/spark-tts-training