MetaVoice-1B (RLX safetensors)

MetaVoice stages converted to safetensors (PyTorch .pt removed).

Field	Value
Hub id	eugenehp/metavoice
Kind	Converted / re-laid-out for RLX from an upstream checkpoint.
RLX crate	rlx-metavoice
Upstream	https://huggingface.co/metavoiceio/metavoice-1B-v0.1

Quick start

hf download eugenehp/metavoice --local-dir .
cargo run -p rlx-metavoice --release -- --model-dir .

File highlights

• first_stage.safetensors (4.7 GiB)
• metavoice.f16.gguf (2.4 GiB)
• second_stage.safetensors (55.2 MiB)
• speaker_encoder.safetensors (5.4 MiB)
• bria_16k.wav (1.4 MiB)
• tokenizer_metavoice.json (7.8 KiB)
• bria_spk_emb.json (3.9 KiB)
• first_stage_args.json (488 B)
• second_stage_args.json (319 B)
• config.json (39 B)

Do not use as the main runtime pack

These files may be present for historical / staging reasons but are not the supported RLX load path:

• metavoice.f16.gguf

Community *.f16.gguf / LM-only Q4 packs are format wraps — they are not drop-in replacements for the RLX primary file above.

Run with RLX

Clone rlx-models, place this repo under weights/tts/metavoice (or pass the path explicitly), then:

cargo run -p rlx-metavoice --release -- --model-dir .

License

Apache License 2.0 — see LICENSE. Inherit upstream terms when redistributing.

Original weights and authorship: https://huggingface.co/metavoiceio/metavoice-1B-v0.1

Maintenance

Cards and LFS attrs are regenerated from the local weights/ tree in rlx-models via python3 scripts/prepare_weights_hf.py.