Fish-Audio-S2-Pro-MLX-fp16

This is an fp16 artifact export of fishaudio/s2-pro for Apple Silicon runtime evaluation.

Runtime support is not implied by this model card. The bundle preserves upstream key names and is intended for a speech-swift Swift/MLX port.

Model

Field	Value
Source	fishaudio/s2-pro
Source revision	1de9996b6be38b745688de084d87a5633f714e4e
Format	MLX fp16 safetensors
License posture	research/non-commercial
Readiness	benchmark-only
Sample rate	See upstream/runtime implementation.
Voice conditioning	Fish Speech reference / speaker conditioning stack
Runtime status	benchmark artifact; not for default product integration

Emotion Control

Field	Value
Marker syntax	free-form inline bracket tags
Supported markers	[pause], [emphasis], [laughing], [excited], [angry], [whisper], [screaming], [shouting], [surprised], [sad]

Files

• config.json - root config for Hugging Face download tracking and runtime metadata
• soniqo_manifest.json - export manifest with source, marker, readiness, and file metadata
• *.safetensors - fp16-converted model weights with upstream key names preserved

Notes

• Strong benchmark for free-form inline prosody/emotion tags.
• Public weights are research/non-commercial; commercial use requires a separate license.
• The codec checkpoint is converted from PyTorch pickle to safetensors when this model is exported.

Usage

These artifacts are for runtime implementation and evaluation. They are not a drop-in Transformers checkpoint.

git clone https://huggingface.co/aufklarer/Fish-Audio-S2-Pro-MLX-fp16

Source

• Upstream model: https://huggingface.co/fishaudio/s2-pro

Links

• speech-swift - Apple SDK
• Speech Studio - local speech generation and voice cloning app
• Docs - install and CLI docs
• soniqo.audio - website
• Blog - blog