export: add FACodec V2 ONNX artifacts (apache-2.0)

PROVENANCE.md

@@ -0,0 +1,25 @@
+# PROVENANCE
+
+| Field | Value |
+|-------|-------|
+| upstream_repo | https://github.com/open-mmlab/Amphion |
+| upstream_ref | main |
+| export_script_sha | 120a3035cced6cb7dc1fc5458efe6efa5ac049db |
+| export_date | 2026-06-12T15:01:04Z |
+| torch_version | 2.10.0+cu128 |
+| onnx_version | 1.20.1 |
+| platform | Linux-6.18.20-1-lts-x86_64-with-glibc2.43 |
+
+## Upstream licence
+
+```
+Apache License
+Version 2.0, January 2004
+http://www.apache.org/licenses/
+
+TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+<...full Apache-2.0 text; see https://www.apache.org/licenses/LICENSE-2.0>
+
+Copyright 2023 Amphion Authors (open-mmlab/Amphion).
+Weights: amphion/naturalspeech3_facodec on Hugging Face (apache-2.0).
+```

README.md

@@ -0,0 +1,70 @@
+---
+license: apache-2.0
+tags:
+- voice-conversion
+- speech
+- onnx
+- audio
+---
+
+# TigreGotico/vconnx-facodec
+
+ONNX artifacts for the FACodec (NaturalSpeech 3) voice-conversion engine,
+part of [vconnx](https://github.com/TigreGotico/vconnx).
+
+## License
+
+Upstream weights: **Apache-2.0** — [amphion/naturalspeech3_facodec](https://huggingface.co/amphion/naturalspeech3_facodec).  
+Code: Amphion (open-mmlab/Amphion) — Apache-2.0.  
+ONNX exports: Apache-2.0 (same upstream license, stated here per vconnx publish-all policy).
+
+## Architecture
+
+FACodec (Ju et al., NaturalSpeech 3, ICML 2024) disentangles speech into
+content, prosody, timbre, and acoustic detail subspaces.  Voice conversion is
+zero-shot: encode source → quantize → swap reference timbre embedding → decode.
+
+| Component | Input → Output | Description |
+|---|---|---|
+|  | wav(1,1,N) → enc_feats(1,256,T) | Convolutional encoder (hop=200) |
+|  | enc_feats(1,256,T) → spk_embs(1,256) | TransformerEncoder timbre extractor |
+|  | (enc_feats, mel_20(1,20,T)) → vq_ids(6,1,T) | Factorised VQ — prosody(1)+content(2)+residual(3) |
+|  | (vq_ids, spk_embs) → wav(1,1,N) | vq2emb + AdaIN + conv decoder |
+
+## Parity (fp32 torch vs ORT)
+
+| Component | max_abs Δ | mean_abs Δ | Verdict |
+|---|---|---|---|
+| encoder | 1.62e-05 | 2.36e-06 | PASS |
+| timbre | 1.43e-06 | 6.40e-08 | PASS |
+| quantize | exact int64 match | — | PASS |
+| decoder | 7.50e-09 | 1.46e-09 | PASS |
+
+## Model sizes
+
+| File | Size |
+|---|---|
+|  (fp32) | 16.5 MB |
+|  (INT8) | 4.7 MB |
+|  (fp32) | 33.0 MB |
+|  (INT8) | 12.1 MB |
+|  (fp32) | 33.4 MB |
+|  (INT8) | 12.5 MB |
+|  (fp32) | 66.2 MB |
+|  (INT8) | 36.8 MB |
+
+## Intelligibility (WER gate ≤ 25%)
+
+Tested with edge-tts source (en-US-GuyNeural) converted to two reference voices
+(en-US-AriaNeural, en-GB-SoniaNeural) using Whisper base.en transcription.
+
+| Reference voice | WER | Gate |
+|---|---|---|
+| en-US-AriaNeural | 0% | ✓ PASS |
+| en-GB-SoniaNeural | 0% | ✓ PASS |
+
+## References
+
+- https://arxiv.org/abs/2403.03100
+- https://huggingface.co/amphion/naturalspeech3_facodec
+- https://github.com/TigreGotico/vconnx

config.json

@@ -0,0 +1,66 @@
+{
+  "engine": "facodec",
+  "components": {
+    "facodec_encoder": "facodec_encoder.onnx",
+    "facodec_encoder_q8": "facodec_encoder_q8.onnx",
+    "facodec_timbre": "facodec_timbre.onnx",
+    "facodec_timbre_q8": "facodec_timbre_q8.onnx",
+    "facodec_quantize": "facodec_quantize.onnx",
+    "facodec_quantize_q8": "facodec_quantize_q8.onnx",
+    "facodec_decoder": "facodec_decoder.onnx",
+    "facodec_decoder_q8": "facodec_decoder_q8.onnx"
+  },
+  "sample_rates": {
+    "output": 16000
+  },
+  "distributable": true,
+  "metadata": {
+    "opset": 14,
+    "upstream_repo": "amphion/naturalspeech3_facodec",
+    "license": "apache-2.0",
+    "n_quantizers": 6,
+    "vq_num_q_p": 1,
+    "vq_num_q_c": 2,
+    "vq_num_q_r": 3,
+    "hop_length": 200,
+    "parity": {
+      "encoder": {
+        "max_abs": 1.621246337890625e-05,
+        "mean_abs": 2.3622947082913015e-06,
+        "pass": true
+      },
+      "timbre": {
+        "max_abs": 1.430511474609375e-06,
+        "mean_abs": 6.403433872037567e-08,
+        "pass": true
+      },
+      "quantize": {
+        "exact_match": true,
+        "pass": true
+      },
+      "decoder": {
+        "max_abs": 7.49787432141602e-09,
+        "mean_abs": 1.4579514262536009e-09,
+        "pass": true
+      }
+    },
+    "sizes_mb": {
+      "facodec_encoder": {
+        "fp32_mb": 16.5,
+        "q8_mb": 4.7
+      },
+      "facodec_timbre": {
+        "fp32_mb": 33.0,
+        "q8_mb": 12.1
+      },
+      "facodec_quantize": {
+        "fp32_mb": 33.4,
+        "q8_mb": 12.5
+      },
+      "facodec_decoder": {
+        "fp32_mb": 66.2,
+        "q8_mb": 36.8
+      }
+    }
+  }
+}

facodec_decoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b49cbf948c06f98fe7a642f3a081c2992276c350da00ab15e0b45b1a63b0bf73
+size 69398668

facodec_decoder_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75316291882b9d14653ffc6da4c3f85ad724fe8a71f40f8096b039cdd09bda69
+size 38604746

facodec_encoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0d327a05d7bc1ec2a53cd6f386b7dd6b0c6c01d23a915eb640403ab44eec0ab
+size 17285976

facodec_encoder_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db6c42b9d5e0ad10353ba5a193d7794ba6dfe039c958571ede56c1eb2777a658
+size 4979562

facodec_quantize.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:197e5f4cb36c3bc3dcb74bd4a4d9ecea6becb685c2bc13ee0058d1ed991763bc
+size 35022902

facodec_quantize_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0db5c5243d08285f4609e623df202936a4bfab04bf1b338ba275b6e49132f8b2
+size 13061686

facodec_timbre.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:865b106d26932fc33f53173cb6a392dd5f9d117a8a51fb9ed592f87508c107ca
+size 34621660

facodec_timbre_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7102d36a9621de25a52a46c9b3ec1e7c731429d7140d7de22de8e57d0dcef73f
+size 12659511