Link to open-source conversion script repo

README.md

@@ -100,6 +100,22 @@ let voice = URL(filePath: "path/to/custom-conds.safetensors")
 let audio = try await model.generate("Hello there.", voice: voice)
 ```
 
+## Conversion script
+
+The script that produced these artifacts from the upstream
+[ResembleAI/chatterbox-turbo](https://huggingface.co/ResembleAI/chatterbox-turbo)
+PyTorch weights is open source:
+
+→ https://github.com/ebrinz/chatterbox-turbo-coreml-converter
+
+It documents the input/output signatures for each file
+(`T3Prefill.mlpackage`, `onnx/language_model_single.onnx`,
+`onnx/conditional_decoder_single.onnx`) and the runtime split used at
+inference time (CoreML for the big prefill, ONNX Runtime for per-step
+decode and the post-T3 audio chain). Use it to reproduce the artifacts
+locally, validate them on your own hardware, or build a consumer for
+them in your own app.
+
 ## License
 
 MIT (same as [Chatterbox Turbo](https://huggingface.co/ResembleAI/chatterbox-turbo))