Document managed four-thread CPU runtime

README.md

@@ -70,9 +70,9 @@ These are held-out text generations, not reconstructions of training audio. Each
 
 No single metric captures TTS quality. Inflect v2 reports **human preference**, **predicted naturalness**, **multi-ASR intelligibility**, **complete footprint**, and **runtime** separately rather than compressing them into one unverifiable score.
 
-| Community preference ↑ | UTMOS22 ↑ | Two-ASR semantic WER ↓ | Complete FP32 weights ↓ |
-| ---: | ---: | ---: | ---: |
-| **66.2%** | **4.395** | **3.99%** | **37.53 MB** |
+| Community preference ↑ | UTMOS22 ↑ | Two-ASR semantic WER ↓ | Complete FP32 weights ↓ | 4-thread CPU throughput ↑ |
+| ---: | ---: | ---: | ---: | ---: |
+| **66.2%** | **4.395** | **3.99%** | **37.53 MB** | **6.28× real-time** |
 
 The headline row always refers to **Inflect-Micro-v2**. Detailed competitor results and protocol boundaries are kept visible below.
 
@@ -131,14 +131,45 @@ transcription errors.
 
 </details>
 
-### 4. Runtime status
+### 4. CPU runtime
 
-The release supports CPU and CUDA inference through the same public API. A
-previous local CPU table has been withdrawn because background host saturation
-made its cross-system throughput figures non-reproducible. A clean-host rerun
-will be published with the complete environment, thread policy, warmup count,
-matched prompts, and raw per-utterance timings. Until then, this card makes no
-headline CPU-speed claim.
+Both Inflect releases synthesize comfortably faster than real time on CPU. The
+managed reference run used a Hugging Face **CPU Upgrade** instance (8 vCPU,
+32 GB RAM) with **four framework threads**, end-to-end text-to-waveform timing,
+and 100 fixed Modern400 prompts. Three complete passes were recorded; the first
+cache-building pass was excluded and the table pools passes two and three.
+
+| Release | Steady-state RTF ↓ | Audio / wall time ↑ |
+| --- | ---: | ---: |
+| **Inflect-Micro-v2** | **0.1593** | **6.28×** |
+| **Inflect-Nano-v2** | **0.0933** | **10.72×** |
+
+These are package-level results from the public PyTorch runtime, not a claim
+that Inflect is the fastest compact TTS system. Hardware, frontend behavior,
+framework, compilation, and thread policy all affect small-model measurements.
+
+<details>
+<summary><strong>Open directional compact-system speed context</strong></summary>
+
+The same managed CPU and four-thread policy were used for a shorter comparator
+pass: the identical 50-prompt prefix, repeated twice. KittenTTS and Piper are
+equal-work pooled across their two tested voices.
+
+| System | Audio / wall time ↑ |
+| --- | ---: |
+| Piper Low | 31.37× |
+| KittenTTS Nano | 13.33× |
+| **Inflect-Nano-v2** | **10.72×** |
+| Supertonic 3 · 3-step | 10.15× |
+| **Inflect-Micro-v2** | **6.28×** |
+| Supertonic 3 · 8-step | 4.37× |
+
+Because Inflect uses the larger 100-prompt steady-state run while comparator
+rows use the shorter 50-prompt confirmation pass, this table is deployment
+context rather than a perfectly matched speed leaderboard. Several comparators
+also use optimized ONNX runtimes while this Inflect release uses PyTorch.
+
+</details>
 
 ### 5. Complete weight footprint
 

docs/EVALUATION.md

@@ -14,10 +14,10 @@ behind those results.
 
 ## Evidence at a glance
 
-| Release | Complete parameters | FP32 weights | UTMOS22 ↑ | Two-ASR semantic WER ↓ | Community preference ↑ |
-| --- | ---: | ---: | ---: | ---: | ---: |
-| **Inflect-Micro-v2** | **9,356,513** | **37.53 MB** | **4.395** | **3.99%** | **66.2%** |
-| **Inflect-Nano-v2** | **3,966,721** | **15.97 MB** | **4.386** | **4.21%** | **63.9%** |
+| Release | Complete parameters | FP32 weights | UTMOS22 ↑ | Two-ASR semantic WER ↓ | Community preference ↑ | 4-thread CPU ↑ |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: |
+| **Inflect-Micro-v2** | **9,356,513** | **37.53 MB** | **4.395** | **3.99%** | **66.2%** | **6.28×** |
+| **Inflect-Nano-v2** | **3,966,721** | **15.97 MB** | **4.386** | **4.21%** | **63.9%** | **10.72×** |
 
 ![Community blind listening](../assets/evidence/human-preference.svg)
 
@@ -161,22 +161,42 @@ balanced tournament or formal MOS.
 
 ## Runtime and throughput
 
-### Current publication status
+### Managed CPU reference
 
-The previous Ryzen 9 3900X comparison is withdrawn. The host was under sustained
-background saturation and the harness did not enforce the same thread policy
-inside every PyTorch and ONNX runtime. Those numbers are not release evidence.
+The published CPU reference uses a Hugging Face **CPU Upgrade** instance
+(8 vCPU, 32 GB RAM) with four framework threads and end-to-end
+text-to-waveform timing. Inflect received 100 fixed Modern400 prompts across
+three complete passes. The first pass populated runtime caches; the published
+steady-state result pools passes two and three.
 
-A replacement CPU benchmark must pass all of these gates before publication:
-
-1. named, idle host with sustained preflight load below the declared threshold;
-2. one isolated process per system;
-3. identical frozen prompt order;
-4. fixed warmup count and fixed thread budget;
-5. explicit PyTorch and ONNX intra/inter-op thread configuration;
-6. end-to-end timing including frontend and waveform generation;
-7. raw per-utterance latency, audio duration, RTF, provider, memory, and host load;
-8. repeated runs with median and dispersion, not one lucky pass.
+| Release | Steady-state RTF ↓ | Audio / wall time ↑ |
+| --- | ---: | ---: |
+| **Inflect-Micro-v2** | **0.1593** | **6.28×** |
+| **Inflect-Nano-v2** | **0.0933** | **10.72×** |
+
+The earlier Ryzen 9 3900X CPU comparison is superseded because background host
+saturation and inconsistent framework thread policies made it unsuitable as
+release evidence.
+
+### Directional compact-system context
+
+The same managed CPU and four-thread policy were used for a shorter comparator
+confirmation: the identical 50-prompt prefix repeated twice. KittenTTS and
+Piper are equal-work pooled across two voices.
+
+| System | Audio / wall time ↑ |
+| --- | ---: |
+| Piper Low | 31.37× |
+| KittenTTS Nano | 13.33× |
+| **Inflect-Nano-v2** | **10.72×** |
+| Supertonic 3 · 3-step | 10.15× |
+| **Inflect-Micro-v2** | **6.28×** |
+| Supertonic 3 · 8-step | 4.37× |
+
+This is deployment context, not a perfectly matched speed leaderboard:
+Inflect's figures come from the larger steady-state run, comparator rows use the
+shorter confirmation pass, and several comparator packages use optimized ONNX
+while the public Inflect package uses PyTorch.
 
 The public runtime also collapses training-time weight normalization once,
 immediately after loading the checkpoint. A fixed-seed regression test produced

release_manifest.json

@@ -89,8 +89,8 @@
     },
     {
       "path": "docs/EVALUATION.md",
-      "bytes": 11854,
-      "sha256": "b2aaf6f467b60d7414f4a4303e09213c382c3e73063deddb76268068dae62730"
+      "bytes": 12599,
+      "sha256": "b350f2990dc6a192211bd065f5f28920cb513507981f42d76e0ad0bc1e6e724b"
     },
     {
       "path": "docs/EXPORTS.md",
@@ -294,8 +294,8 @@
     },
     {
       "path": "README.md",
-      "bytes": 17198,
-      "sha256": "006177225cb135e5f376f7ac7383a93569d30639d3e555c2dfb500d84a04f7d5"
+      "bytes": 18551,
+      "sha256": "a5ed283e363b362cebce9576a2e9d2b29ac2989d34e391d9b29d060b1b5170b6"
     },
     {
       "path": "requirements-tested.txt",