Resonance step 20000 — weights-only

README.md

@@ -1,25 +1,20 @@
 ---
-license: apache-2.0
+license: other
 tags: [resonance, anima, tts, speech, from-scratch]
-library_name: fsp-ttv
 ---
 
-# FatStinkyPanda · Resonance — step 5,000
+# FatStinkyPanda · Resonance — step 20,000
 
-A from-scratch **neural text-to-speech** engine (~309M params, SNAC 24 kHz codec), trained in the **Anima** project (https://github.com/FatStinkyPanda/Anima). Created & released by **FatStinkyPanda** (Daniel A. Bissey).
+A from-scratch **neural text-to-speech** engine (SNAC 24 kHz codec), trained in the **Anima** project (https://github.com/FatStinkyPanda/Anima). The published `resonance.safetensors` holds the trained tensors; the architecture loads them via the project's private engine.
 
-> Honest scope: a custom architecture trained on a tiny budget. This repo holds the raw checkpoint — it loads with the project's `fsp-ttv` engine, NOT `transformers`.
+> **Proprietary architecture.** The architecture and training innovations are NOT public. This repo hosts the trained **weights + tokenizer only**; the config + modeling code stay private. Benchmarks run on **neutral free compute** and are published openly so the numbers are verifiable without exposing how the model works.
 
-## Synthesize (fsp-ttv engine)
-```bash
-pip install 'git+https://github.com/FatStinkyPanda/Anima#subdirectory=engine'
-fsp-ttv synthesize "Hello from Resonance." --ckpt final.pt -o out.wav
-```
-
-## Latest validation health scores (0..1, from the engine's own instrument)
+## Latest held-out benchmarks (neutral compute)
 - **behavior**: 1.0
 - **layer_health**: 1.0
 - **stability**: 1.0
-- **coarse_accuracy**: 0.0042
-- **overall**: 0.6384
-- **tf_coarse_accuracy**: 0.0042
+- **coarse_accuracy**: 0.0526
+- **overall**: 0.6487
+- **tf_coarse_accuracy**: 0.0526
+
+Created & released by **FatStinkyPanda**. For access or collaboration, contact the creator.