Upload VoiceAcousticTransfer/README.md with huggingface_hub
Browse files
VoiceAcousticTransfer/README.md
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# VoiceAcousticTransfer — P2a noise_flow (2026-07-09)
|
| 2 |
+
|
| 3 |
+
- `p2a_noise_flow_scaled_lufs20_40k_best_step40000.pt` — recording-condition transfer model
|
| 4 |
+
(conditional flow matching, hidden 1024 / depth 16 / heads 16, 40k steps, LUFS-normalized
|
| 5 |
+
raw-DAC caches). Optimizer stripped; contains model + EMA + all metadata incl.
|
| 6 |
+
`latent_scale` (1.652), `daps_to_idx`, `style_to_idx`.
|
| 7 |
+
Inference: Euler t 0->1, 32 steps, joint-axis CFG 2.0; inputs loudness-normalized to -20 LUFS.
|
| 8 |
+
- `trackb_condition_analyzer_best.pt` — standalone recording-condition analyzer
|
| 9 |
+
(contrastive encoder + a18/recipe-parameter regression + DAPS classifier heads).
|
| 10 |
+
- `condition_gmm.joblib` — in-the-wild condition prior (GMM over analyzer embeddings).
|
| 11 |
+
|
| 12 |
+
Code, evaluation harness, and usage: https://github.com/Kiberchaika/VoiceAcousticTransfer
|
| 13 |
+
(branch `rec-con-diffusion`); run log `8jul_acoustictransfer_evolution_report_journal.md`.
|