feat(kokoro): train and verify bilingual PyTorch weights

models/tts/kokoro-v1.1-zh/AGENTS.md

@@ -5,7 +5,7 @@ planning changes, asking product-scope questions, or starting model execution.
 It contains the shared requirements, artifact gaps, gate sequence, and runbook.
 
 - Preserve settled scope: one female voice, English/Mandarin/code-switching,
-  compact Kokoro v1.1-zh-compatible adaptation, FluidAudio/Core ML direction;
+  compact Kokoro v1.1-zh-compatible PyTorch adaptation; Core ML is deferred;
   no voice cloning or runtime multi-speaker expansion.
 - Inspect the receiving environment's existing setup and approvals first. Ask
   only for genuinely missing inputs; do not make the user repeat settled goals.

models/tts/kokoro-v1.1-zh/training/.gitignore

@@ -0,0 +1,4 @@
+.artifacts/
+.runs/
+.pytest_cache/
+*.egg-info/

models/tts/kokoro-v1.1-zh/training/LICENSE.kokoro

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

models/tts/kokoro-v1.1-zh/training/MODEL_CARD.md

@@ -0,0 +1,102 @@
+# Kokoro MF5 English/Mandarin adaptation
+
+## Artifact
+
+A single female voice adapted from Kokoro-82M-v1.1-zh, retaining its architecture,
+vocabulary, 24 kHz output, and length-indexed 256-dimensional style interface.
+The selected checkpoint is deterministic run v2, update **500**. Weights remain
+local, along with generated evaluation audio.
+
+| File | Bytes | SHA-256 |
+|---|---:|---|
+| `model.pth` | 327,453,363 | `2c27865b7c794124197e49ab9b7c814da088e9ef87f1dd605f95b87e1458a4c0` |
+| `voice.pt` | 523,739 | `119eca879d472a5327bc1e01c358958e9028a62d09d3a528d953248e94fbc8b3` |
+
+Use the matching voice table, configuration, and frontend with these weights.
+[results.json](results.json) records model/data/report identities and exact counts.
+
+## Data and training
+
+Real EMIME MF5 microphone-0 recordings: 314 utterances, approximately 28 minutes.
+Official test IDs and translated parallel passages stay together. Other passages
+have deterministic development assignment. There is one session per language,
+so these splits are passage-disjoint, not session-disjoint.
+
+| Split | English utterances | Mandarin utterances | Prepared seconds |
+|---|---:|---:|---:|
+| Train | 96 | 129 | 1,347.95 |
+| Development | 6 | 9 | 83.15 |
+| Test | 43 | 31 | 274.55 |
+
+The frontend uses pinned Misaki, NFKC and simplified Chinese, explicit API/GitHub
+pronunciations, and distinct numeral/erhua handling. Unknown symbols and overly
+long inputs fail. Audio is resampled from 22,050 to 24,000 Hz and padded to the
+600-sample duration grid. Temporary baseline synthesis supplies token boundaries
+transferred by MFCC DTW onto real recordings. This is approximate alignment;
+**only real recordings are reconstruction targets**. Pinned JDC supplies F0;
+pitch/energy targets use the 300-sample grid and StyleTTS2 mel convention.
+
+Trainable parameters: ALBERT projection, prosody predictor, text encoder, and
+one shared voice-style residual. ALBERT, decoder, and identity affine constants
+are frozen. AdamW uses LR 2e-5 (style: 1e-3), two accumulated utterances per step,
+2.4-second crops, balanced language sampling, and seed 1729. Loss combines mel,
+duration, F0, energy, and style regularization. Strict deterministic execution
+uses equivalent slice-based reflection padding to avoid nondeterministic CUDA
+padding gradients. Checkpoints retain optimizer and all RNG states.
+
+The eight-recording overfit check reduced loss 43.1%. After an exploratory
+1,000-update run, the corrected deterministic run started from the base weights
+and completed 500 updates. Its development acoustic loss fell 5.98555 → 3.19828.
+Candidates 100/300/500 were compared on development ASR; 300 and 500 tied,
+with lower acoustic loss selecting 500. Test results did not influence selection.
+
+## Measured quality
+
+Matched baseline/candidate inputs, seed, speed, frontend, and respective voice
+tables. Whisper large-v3-turbo uses pinned weights, greedy decoding, and explicit
+English/Chinese transcription. Scores apply NFKC, simplified Chinese, lowercase,
+and punctuation removal; numeral spellings are not equated. Evaluation uses
+FP32 generator tensors, FP16 ASR, matmul TF32 off, and cuDNN TF32 allowed.
+
+| Metric | Baseline | Trained |
+|---|---:|---:|
+| Development English WER, 53 words | 1.89% | 0.00% |
+| Development Mandarin CER, 235 characters | 7.23% | 5.96% |
+| Held-out English WER, 325 words / 43 utterances | 6.15% | 4.31% |
+| Held-out Mandarin CER, 466 characters / 31 utterances | 1.50% | 1.93% |
+| Mixed-control raw CER, 6 utterances | 13.18% | 19.38% |
+
+English ASR improved; Mandarin test CER increased by two character errors.
+The mixed score regressed under a recognizer that sometimes translates mixed
+speech instead of transcribing it. It is retained as a diagnostic, not treated
+as a reliable measurement of code-switch quality. ASR does not establish
+naturalness, Mandarin tone correctness, or speaker consistency. No blinded
+bilingual listening study has been completed.
+
+## Implementation validation
+
+- Untouched baseline: 688 state tensors and 15 inference cases match upstream.
+- Eight real-recording gradient/forward checks pass; alignment negative controls
+  distinguish matching text from wrong-text recordings.
+- Three updates uninterrupted versus save/restart after update one match
+  bit-for-bit, including model, optimizer, RNG, and validation state.
+- Exported weights/voice match the selected checkpoint; upstream Kokoro produces
+  identical waveforms/durations on 12 controls. Exactly 148 generator tensors
+  changed; frozen weights did not change.
+- Standalone offline CPU and GPU synthesis pass. Peak training allocation: 2.01 GB.
+
+## Limits and provenance
+
+This is an experimental model. The corpus provides no real code-switch speech,
+no session-disjoint validation, and no independently certified production voice
+consent. Listening, tone, and naturalness acceptance remain open. No deployment
+or Core ML acceptance is implied.
+
+Corpus: [EMIME](https://www.emime.org/participate/emime-bilingual-database.html)
+(ODbL/DbCL per its README). Base code/weights and auxiliary models have separate
+licenses; see [NOTICE.md](NOTICE.md). Acquisition revisions and checksums are in
+[baseline.lock.json](baseline.lock.json) and [training-assets.lock.json](training-assets.lock.json).
+The original full run reports remain in
+[Git history](https://github.com/FluidInference/mobius/tree/fb13fc48d599b733759a4ecc60d0c10b6499fffc/models/tts/kokoro-v1.1-zh/training/evidence)
+and the local ignored run directories; the recorded run's dependency-lock hash
+is retained in `results.json`.

models/tts/kokoro-v1.1-zh/training/NOTICE.md

@@ -0,0 +1,30 @@
+# Upstream implementation attribution
+
+The compatible generator assembly and inference orchestration follow
+[hexgrad/kokoro](https://github.com/hexgrad/kokoro), commit
+`dfb907a02bba8152ca444717ca5d78747ccb4bec`, especially `kokoro/model.py`.
+Kokoro's source is Apache-2.0 licensed; see [LICENSE.kokoro](LICENSE.kokoro).
+The actual layer modules are installed from that pinned upstream source,
+including its StyleTTS2/iSTFTNet attribution. They are not vendored here.
+
+The model checkpoint, voice data, training-corpus candidates, and source code
+have separate provenance and use requirements. This file does not grant new
+rights to recordings or a derived production voice.
+
+Acoustic target extraction follows the published StyleTTS2 training convention
+and loads the JDC pitch model from
+[yl4579/StyleTTS2](https://github.com/yl4579/StyleTTS2/tree/5cedc71c333f8d8b8551ca59378bdcc7af4c9529),
+pinned in `training-assets.lock.json`. Acquisition downloads its MIT license
+alongside the target model. The JDC source/checkpoint is not redistributed in Git.
+
+EMIME Mandarin/English recordings are credited to Mirjam Wester, Hui Liang,
+and the Centre for Speech Technology Research, University of Edinburgh.
+The corpus archive's own README specifies ODbL 1.0 and DbCL 1.0. Preserve its
+README/license with acquired data. This experiment uses MF5 microphone 0;
+the toolkit does not grant publicity/personality rights or certify consent
+for a production synthetic voice.
+
+The evaluation ASR model is OpenAI Whisper large-v3-turbo (MIT). Misaki and
+its language resources, eSpeak, OpenCC, spaCy, and other installed dependencies
+retain their respective licenses. `uv.lock` records the actual packages;
+bundling a trained checkpoint does not relicense those components.