Irodori-TTS-MLX-500M-v2-VoiceDesign

Irodori-TTS-MLX-500M-v2-VoiceDesign

Converted MLX RF-DiT weights for Aratako/Irodori-TTS-500M-v2-VoiceDesign, produced for the unofficial t0yohei/Irodori-TTS-MLX runtime.

This is a converted MLX artifact, not the original upstream checkpoint. It is published as an approved hosted RF-DiT weights layout for the v0.2 irodori-tts-mlx runtime contract.

Provenance

• Upstream checkpoint: Aratako/Irodori-TTS-500M-v2-VoiceDesign
• Upstream revision: 456e55708e7183f5c7faa1448209d54aa8991451
• Source file: model.safetensors
• Converted with: t0yohei/Irodori-TTS-MLX commit 76bec1bd1c98d12698892b3ccd050b703c199a61
• Approved hosted artifact revision: bf877a3beb7d921dc6bfb2b6812d02be07f39f2a
• Detected family: voicedesign

Access

This repository is gated. Accept the repository conditions on Hugging Face and use an authenticated huggingface_hub session before resolving it with --weights-repo.

Usage

Install the runtime and make the upstream Irodori-TTS package importable as described in the Irodori-TTS-MLX README. The hosted repository path also requires huggingface_hub.

PYTHONPATH=/path/to/Irodori-TTS:${PYTHONPATH:-} \
irodori-tts-generate \
  --weights-repo t0yohei/Irodori-TTS-MLX-500M-v2-VoiceDesign \
  --weights-revision bf877a3beb7d921dc6bfb2b6812d02be07f39f2a \
  --text "こんにちは。今日は良い天気です。" \
  --caption "落ち着いた女性の声" \
  --no-reference \
  --output /tmp/irodori-hosted.wav \
  --preset balanced \
  --json

VoiceDesign supports caption-conditioned no-reference generation. Omit --seconds for the normal bounded text-length fallback, or pass --seconds only when a specific prompt needs a manual duration override.

By default, the current irodori-tts-mlx runtime resolves the separate approved MLX DACVAE codec artifact for encode/decode. RF-DiT weights and DACVAE codec artifacts intentionally stay in separate repositories. Pin the codec contract with --codec-runtime-mode mlx, --codec-artifact-repo t0yohei/Irodori-TTS-MLX-DACVAE-Codec, and --codec-artifact-revision bb89840af0deb729cc7a8e4ba5ebddb49e2b3e78 when reproducibility matters.

Artifact Boundary

This repository includes the converted RF-DiT/text/conditioning weights and lightweight loader metadata for irodori-tts-mlx. It does not bundle Semantic-DACVAE codec weights, upstream source code, reference audio, generated audio samples, Hugging Face cache snapshots, training data, tokenizer assets, or unaudited model artifacts. The included tokenizer_config.json is metadata for the runtime contract, not a redistributed tokenizer asset.

The current public generation path uses the separate approved MLX DACVAE codec artifact instead of the removed upstream PyTorch DACVAE bridge. This repository does not bundle that codec artifact; resolve it with the runtime default, --codec-artifact-repo, or a local --codec-artifact-dir.

Users remain responsible for following the upstream model card, applicable law, and any rights attached to prompts, reference audio, and generated audio. The upstream model card includes ethical-use restrictions, including no impersonation and no misleading synthetic speech.

See the redistribution audit: https://github.com/t0yohei/Irodori-TTS-MLX/blob/main/docs/preconverted_weights_redistribution_audit.md

For the complete hosted-layout contract and local conversion fallback, see: https://github.com/t0yohei/Irodori-TTS-MLX/blob/main/docs/hosted_weights_usage.md