# Tabi WD audition — September 6, 2026

Status at page generation: 16 ready. Playable local clips: 16/16.

Six new Orpheus Q8 Tara clips from eimu use the same expression, laughter,
chuckle/sigh, ordinary, conversation, and technical passages as these demos,
with native Orpheus event tags. They are outside the 16-clip Tabi run count
and event manifest. See [the Orpheus emotion manifest](../orpheus/emotions-20260906/manifest.json).
Its full HTTP request timings include queueing and transfer and are not isolated
generation times. Four older Tabi Q8/Q4 benchmark clips remain in a collapsed
section with [their original manifest](../orpheus/manifest.json).

This page is built from transferred JSONL records and WAVs. The generator neither
contacts Tabi nor runs inference. Missing recordings remain visibly pending.

Tabi's previously verified hardware: Intel Core Ultra 7 258V, 8 CPU threads,
Intel Arc 140V / Xe2, 32 GB shared memory; Gentoo live USB clone of eimu.
New-run environment records below are the authority for actual execution settings.
WD refers to the requested model storage disk; it does not imply GPU acceleration.
The disk path and installed model revisions are not inferred from that request.

Generation time is full waveform generation after model loading and voice
conditioning; RTF = generation seconds / audio seconds. This is batch inference,
not measured first-audio latency or end-to-end network/playback latency. Real-time
generation is not a requirement. Playback uses original WAVs with no new loudness
matching, pitch changes, or other processing. Listen to judge laughter and quality.

Azelma and Eponine lead because the earlier reference provenance identifies them
as female. Alba and Tara-reference are also included. Tara-reference is synthetic
Orpheus reference audio conditioning Chatterbox. No subjective quality ranking is
claimed. The separately labelled Orpheus section contains actual saved Orpheus
output; Tara-reference in the new-run section remains Chatterbox output.

## Captured environment records

```json
[
  {
    "model": "nano",
    "model_repo": "ResembleAI/chatterbox-nano",
    "model_revision": "71ccd1d0081b430592cea481f4307e764e07bc64",
    "host": "tabi",
    "seed": 42,
    "torch_threads": 4,
    "device": "cpu",
    "timestamp": "2026-09-06T07:09:24.626535+00:00",
    "phase": "environment",
    "version": "0.1.7",
    "torch": "2.6.0+cpu",
    "python": "3.11.15",
    "platform": "Linux-7.0.3-x86_64-Intel-R-_Core-TM-_Ultra_7_258V-with-glibc2.43",
    "processor": "Intel(R) Core(TM) Ultra 7 258V",
    "memory_bytes": 33073115136,
    "work_dir": "/mnt/navi-ssd-bench/Navi-Voice/runs/20260906",
    "cache_root": "/mnt/navi-ssd-bench/Navi-Voice/models/hf-cache",
    "storage_root": "/mnt/navi-ssd-bench",
    "disk_label": "WD WDS100T2B0A serial21313F802709 UUID592C-BEA3",
    "snapshot": "/mnt/navi-ssd-bench/Navi-Voice/models/hf-cache/hub/models--ResembleAI--chatterbox-nano/snapshots/71ccd1d0081b430592cea481f4307e764e07bc64",
    "harness_sha256": "76ebd85b515dd3151183d7a0b026c94ebebbc2931dd438200fbeb774850599aa",
    "voices": [
      "azelma",
      "eponine",
      "tara-reference",
      "alba"
    ],
    "passages": {
      "ordinary": "Hey Aku. The laptop found its voice. That little victory made my day.",
      "laugh": "Hey Aku. [laugh] The laptop found its voice. [chuckle] That little victory made my day."
    },
    "note": "New short paired samples. No speed cutoff. First sample may warm up."
  },
  {
    "model": "turbo",
    "model_repo": "ResembleAI/chatterbox-turbo",
    "model_revision": "749d1c1a46eb10492095d68fbcf55691ccf137cd",
    "host": "tabi",
    "seed": 42,
    "torch_threads": 4,
    "device": "cpu",
    "timestamp": "2026-09-06T07:11:08.636099+00:00",
    "phase": "environment",
    "version": "0.1.7",
    "torch": "2.6.0+cpu",
    "python": "3.11.15",
    "platform": "Linux-7.0.3-x86_64-Intel-R-_Core-TM-_Ultra_7_258V-with-glibc2.43",
    "processor": "Intel(R) Core(TM) Ultra 7 258V",
    "memory_bytes": 33073115136,
    "work_dir": "/mnt/navi-ssd-bench/Navi-Voice/runs/20260906",
    "cache_root": "/mnt/navi-ssd-bench/Navi-Voice/models/hf-cache",
    "storage_root": "/mnt/navi-ssd-bench",
    "disk_label": "WD WDS100T2B0A serial21313F802709 UUID592C-BEA3",
    "snapshot": "/mnt/navi-ssd-bench/Navi-Voice/models/hf-cache/hub/models--ResembleAI--chatterbox-turbo/snapshots/749d1c1a46eb10492095d68fbcf55691ccf137cd",
    "harness_sha256": "76ebd85b515dd3151183d7a0b026c94ebebbc2931dd438200fbeb774850599aa",
    "voices": [
      "azelma",
      "eponine",
      "tara-reference",
      "alba"
    ],
    "passages": {
      "ordinary": "Hey Aku. The laptop found its voice. That little victory made my day.",
      "laugh": "Hey Aku. [laugh] The laptop found its voice. [chuckle] That little victory made my day."
    },
    "note": "New short paired samples. No speed cutoff. First sample may warm up."
  }
]
```

## Files

- [All transferred event records](manifest.json)
- [Previous saved laughter audition](../laughter/index.html): M1 Air measurements
- [Previous feminine report](../feminine/REPORT.md): M1 Air measurements
- [Previous reference provenance](../feminine/reference-provenance.json)

Regenerate locally after transferring JSONL and raw/ WAVs into this directory:

```sh
python benchmarks/tabi-wd-20260906/make_page.py
```

Full HTML path: `/home/akujin/Sources/Navi-Voice/website/www/tabi-wd-20260906/index.html`.
The static website server is not started by this generator.