Add verified Irodori-TTS-v4.1-Small iOS ONNX pack

LICENSES/ModernBERT-MIT.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 SB Intuitions
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -14,6 +14,7 @@ tags:
 
 ONNX model packs exported for the Onsei iOS client from the following upstream models:
 
+- [`Aratako/Irodori-TTS-v4.1-Small`](https://huggingface.co/Aratako/Irodori-TTS-v4.1-Small), source revision `2b28324dc263ed5e6638b3cf3dd94c82ead07b4b`
 - [`Aratako/Irodori-TTS-500M-v3`](https://huggingface.co/Aratako/Irodori-TTS-500M-v3), source revision `236c1e56591279fc24e3c1bf6609fc06e48dde28`
 - [`Aratako/Irodori-TTS-600M-v3-VoiceDesign`](https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign), source revision `e863a3a93e652e09afeff3e84823a206a0a60314`
 - [`Aratako/Semantic-DACVAE-Japanese-32dim`](https://huggingface.co/Aratako/Semantic-DACVAE-Japanese-32dim)
@@ -23,6 +24,7 @@ These files are converted runtime artifacts, not new model weights. They are not
 
 ## Packs
 
+- `v4.1-small/`: unified Japanese TTS, reference cloning and VoiceDesign, with the improved v4.1 duration predictor. `text_encoder.onnx` contains the shared ModernBERT backbone and both `text_state` and `caption_state` outputs. The app reuses one session for both inputs. No separate `caption_encoder.onnx` is needed. Requires the v4.1-compatible iOS client.
 - `500m-v3/`: standard Japanese TTS and reference-audio voice cloning.
 - `600m-v3-vd/`: VoiceDesign model with caption conditioning.
 
@@ -30,7 +32,7 @@ Each pack contains the split `context_kv.onnx` + `dit_step.onnx` execution path
 
 ## Licenses and attribution
 
-The Irodori-TTS model weights, Semantic-DACVAE model weights, and conversion code are distributed under the MIT License. The llm-jp tokenizer is distributed under Apache License 2.0. See `LICENSES/` and `THIRD_PARTY_NOTICES.md`. Each component remains subject to its own license; no relicensing is implied.
+The Irodori-TTS model weights, Semantic-DACVAE model weights, ModernBERT-ja-310m encoder/tokenizer, and conversion code are distributed under the MIT License. The legacy llm-jp tokenizer is distributed under Apache License 2.0. See `LICENSES/` and `THIRD_PARTY_NOTICES.md`. Each component remains subject to its own license; no relicensing is implied.
 
 ## Responsible use
 
@@ -54,4 +56,3 @@ The ONNX export used here does not add a SilentCipher watermark. Applications us
   howpublished = {https://huggingface.co/Aratako/Irodori-TTS-500M-v3}
 }
 ```
-

THIRD_PARTY_NOTICES.md

@@ -2,6 +2,8 @@
 
 ## Irodori-TTS
 
+- v4.1 source: https://huggingface.co/Aratako/Irodori-TTS-v4.1-Small (revision `2b28324dc263ed5e6638b3cf3dd94c82ead07b4b`)
+- v4 implementation: https://github.com/Aratako/Irodori-TTS (revision `8224dafb46d0aba89209a8f905f1cb7e3299d9c1`); shared backbone export and tensor-based attention mask handling for dynamic ONNX inputs.
 - Copyright: Chihiro Arata and contributors
 - Sources: https://huggingface.co/Aratako/Irodori-TTS-500M-v3 and https://huggingface.co/Aratako/Irodori-TTS-600M-v3-VoiceDesign
 - License: MIT (`LICENSES/MIT.txt`)
@@ -28,3 +30,9 @@ The responsible-use restrictions and disclaimers from the upstream model cards a
 - Source: https://github.com/Aratako/Irodori-TTS
 - License: MIT (`LICENSES/MIT.txt`)
 
+## ModernBERT-ja-310m (v4.1 encoder and tokenizer)
+
+- Source: https://huggingface.co/sbintuitions/modernbert-ja-310m
+- Copyright: SB Intuitions and contributors
+- License: MIT
+- The fine-tuned encoder and exact tokenizer are redistributed as part of the Irodori-TTS-v4.1-Small conversion. Text and caption share one encoder graph.

manifest.json

@@ -1,7 +1,7 @@
 {
   "schemaVersion": 1,
   "repository": "raratu/Onsei-iOS-Models",
-  "revision": "ios-v1",
+  "revision": "ios-v2",
   "packs": {
     "500m-v3": {
       "files": [
@@ -105,6 +105,60 @@
           "sha256": "955dc1fa623fab38cc92a3f4ee172423ae6d73201c4207569bfdf5626bc733f0"
         }
       ]
+    },
+    "v4.1-small": {
+      "files": [
+        {
+          "path": "config.json",
+          "size": 2071,
+          "sha256": "f6b12d8c6b076b08fea876d402008e659773f2300c0f1e279b3b5174e968535c"
+        },
+        {
+          "path": "text_encoder.onnx",
+          "size": 1272933729,
+          "sha256": "961efbf4b01e47114e75b62210f29db1568acf244308c7b8657c245061fc2f39"
+        },
+        {
+          "path": "context_kv.onnx",
+          "size": 220370311,
+          "sha256": "5700d738f68ac9df1ed868f7fa983450897a240531d4877c4794e1ba9087c22e"
+        },
+        {
+          "path": "dit_step.onnx",
+          "size": 1243780959,
+          "sha256": "e772e50cdfbc2c72d6ea153cbcbe5f00ec10d4006d9bc1bc727e4a6f69b21877"
+        },
+        {
+          "path": "duration.onnx",
+          "size": 87172678,
+          "sha256": "b807f0e9a3e86bce4fd4324db6da2ded1627f6af3ed155b918e249f2fde50809"
+        },
+        {
+          "path": "speaker_encoder.onnx",
+          "size": 242880837,
+          "sha256": "ecfa3a47f0a5bde19a4fcdca46d27fe51da97fbf68897476dca313017931945f"
+        },
+        {
+          "path": "dacvae_encode.onnx",
+          "size": 109452758,
+          "sha256": "3ccc52d3a5b7af5a7f2ade7f187e277c6f7d457b609db2206ea25b8956b34169"
+        },
+        {
+          "path": "dacvae_decode.onnx",
+          "size": 261400162,
+          "sha256": "48f55372b800130ce3f56895501fcb154bac486a9ac805da1c39ccd06e2d1e48"
+        },
+        {
+          "path": "tokenizer/tokenizer.json",
+          "size": 6718495,
+          "sha256": "6a0734cf21c802169defaffe719bc2ef12bb9d0be37e54b61ed27aa89394723d"
+        },
+        {
+          "path": "tokenizer/tokenizer_config.json",
+          "size": 668,
+          "sha256": "d229a271c64de1a7939d20d3665498e873fa91d5ee2edf135d73ec752cb9c9d3"
+        }
+      ]
     }
   }
 }

v4.1-small/config.json

@@ -0,0 +1,72 @@
+{
+  "model": {
+    "latent_dim": 32,
+    "latent_patch_size": 1,
+    "model_dim": 1280,
+    "num_layers": 12,
+    "num_heads": 20,
+    "mlp_ratio": 2.875,
+    "text_mlp_ratio": 2.6,
+    "speaker_mlp_ratio": 2.6,
+    "dropout": 0.0,
+    "text_vocab_size": 102400,
+    "text_tokenizer_repo": "sbintuitions/modernbert-ja-310m",
+    "text_encoder_revision": "77675fc96a7e445e982e2ba90246b816efc74ec6",
+    "text_add_bos": true,
+    "text_encoder_type": "pretrained",
+    "pretrained_projector_type": "residual_mlp",
+    "pretrained_projector_hidden_ratio": 2.0,
+    "pretrained_projector_dropout": 0.0,
+    "text_dim": 512,
+    "text_layers": 14,
+    "text_heads": 10,
+    "use_caption_condition": true,
+    "use_speaker_condition": true,
+    "caption_vocab_size": 102400,
+    "caption_tokenizer_repo": "sbintuitions/modernbert-ja-310m",
+    "caption_add_bos": true,
+    "caption_dim": 512,
+    "caption_layers": null,
+    "caption_heads": null,
+    "caption_mlp_ratio": null,
+    "speaker_dim": 768,
+    "speaker_layers": 8,
+    "speaker_heads": 12,
+    "speaker_patch_size": 4,
+    "timestep_embed_dim": 512,
+    "adaln_rank": 192,
+    "norm_eps": 1e-05,
+    "use_duration_predictor": true,
+    "duration_aux_dim": 14,
+    "duration_hidden_dim": 1024,
+    "duration_layers": 3,
+    "duration_dropout": 0.1,
+    "duration_attention_heads": 8,
+    "duration_architecture": "token_sum_dual_adarn_zero_no_aux",
+    "duration_token_init_frames": 9.0,
+    "duration_speaker_fusion": "adarn_zero",
+    "duration_caption_fusion": "adarn_zero",
+    "duration_caption_pooling": "masked_mean"
+  },
+  "inference": {
+    "max_text_len": 256,
+    "max_caption_len": 512,
+    "ref_max_seconds": 120.0
+  },
+  "export": {
+    "seq_len": 128,
+    "text_len": 256,
+    "caption_len": 512,
+    "ref_len": 200,
+    "sample_rate": 48000,
+    "hop_length": 1920,
+    "requires_duration": true,
+    "shared_condition_encoder": true,
+    "context_kv_cache": true,
+    "latent_dim": 32,
+    "num_layers": 12,
+    "has_caption": true,
+    "has_speaker_kv_scale": true,
+    "lora_merged": false
+  }
+}

v4.1-small/context_kv.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5700d738f68ac9df1ed868f7fa983450897a240531d4877c4794e1ba9087c22e
+size 220370311

v4.1-small/dacvae_decode.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48f55372b800130ce3f56895501fcb154bac486a9ac805da1c39ccd06e2d1e48
+size 261400162

v4.1-small/dacvae_encode.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ccc52d3a5b7af5a7f2ade7f187e277c6f7d457b609db2206ea25b8956b34169
+size 109452758

v4.1-small/dit_step.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e772e50cdfbc2c72d6ea153cbcbe5f00ec10d4006d9bc1bc727e4a6f69b21877
+size 1243780959

v4.1-small/duration.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b807f0e9a3e86bce4fd4324db6da2ded1627f6af3ed155b918e249f2fde50809
+size 87172678

v4.1-small/speaker_encoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecfa3a47f0a5bde19a4fcdca46d27fe51da97fbf68897476dca313017931945f
+size 242880837

v4.1-small/text_encoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:961efbf4b01e47114e75b62210f29db1568acf244308c7b8657c245061fc2f39
+size 1272933729

v4.1-small/tokenizer/tokenizer_config.json

@@ -0,0 +1,25 @@
+{
+  "add_dummy_prefix_space": false,
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "<cls>",
+  "do_lower_case": false,
+  "eos_token": "</s>",
+  "extra_ids": 0,
+  "is_local": false,
+  "keep_accents": true,
+  "legacy": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "padding_side": "right",
+  "sep_token": "<sep>",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}