{
 "schema_version": "1.0",
 "page_kind": "article",
 "pack_kind": "generate",
 "slug": "listen-for-yourself",
 "title": "Listen for Yourself",
 "dek": "Every clip of the fourth adapter’s four rounds — sixty-five renders, raw and brought to a common loudness — so you can hear what one listener heard and decide for yourself.",
 "published": "2026-09-03",
 "series": [
  "guide"
 ],
 "licence": "CC BY 4.0",
 "status": "pending-judge",
 "pack_note": "no judge is seated, so every chip is held",
 "notice": "items with published:false are held and are not this page's published words",
 "source": {
  "url": "https://research.strata2signal.com/listen-for-yourself/",
  "md_url": "https://research.strata2signal.com/listen-for-yourself/index.md",
  "md_sha": "39e4818280cec4dc01ab14cd7859cf4d3a74e77435c4562fac9b46b71bb4cbe9",
  "html_sha": "492d070b12a5be050cd5f9ac7b36b55cbd81f19cba0b227fbc3138934b8d229c",
  "anchors_sha": "f81a13c7807959e1f56651e3b5648f8639ac6ad3fd5227485627b52a7cbf6860"
 },
 "short": {
  "paragraph": "Every clip of the fourth adapter's four rounds - sixty-five renders, raw and brought to a common loudness - so you can hear what one listener heard and decide for yourself.",
  "from_dek": true,
  "counts": {
   "words": 5660,
   "minutes": 26,
   "tables": 17,
   "kit": false
  },
  "bullets": [
   {
    "text": "the raw renders sit between −19.65 to −13.29 LUFS",
    "figure": "-19.65 to -13.29 LUFS",
    "cite": "how-to-listen",
    "quote": "The raw renders run from -19.65 to -13.29 LUFS integrated - 6.4 units, or about two notches on a fader - so playing twenty in a row means working the volume knob rather than comparing adapters.",
    "span": [
     3418,
     3440
    ]
   },
   {
    "text": "the probe was rendered on 2026-08-31",
    "figure": "2026-08-31",
    "cite": "round-two-the-caption-shape-probe",
    "quote": "Round two - the caption-shape probe The probe that followed the same night, rendered 2026-08-31 between 03:23 and 03:27 UTC: the two cheapest explanations for the verdict, tested with seven more sixty-second clips at the same pins.",
    "span": [
     8200,
     8211
    ]
   },
   {
    "text": "the due-diligence round was rendered on 2026-09-03",
    "figure": "2026-09-03",
    "cite": "round-four-twelve-requests-four-ways",
    "quote": "Round four - twelve requests, four ways The due-diligence round, rendered 2026-09-03 between 03:13 and 03:30 UTC after one of us asked whether the verdict depended on the request: twelve requests written in advance, from the corpus's own genre outward, each with its own tempo, A minor throughout, thirty seconds each, and each rendered four ways on one seed - nothing added, our adapter at half strength, our adapter at full, the one-artist adapter at full.",
    "span": [
     13279,
     13290
    ]
   }
  ]
 },
 "sections": [
  {
   "id": "how-to-listen",
   "heading": "How to listen",
   "level": 2,
   "span": [
    1135,
    5022
   ],
   "chunks": [
    [
     1135,
     5022
    ]
   ],
   "chars": 3887,
   "digest": "the page plays listening masters of raw renders from ACE-Step 1.5's fast variant at eight steps. masters are created by applying one constant gain to reach −16 LUFS or a −1 dBTP ceiling without limiting. tables provide integrated loudness in LUFS, true peak, loudness range, sha256 fingerprints, gain in dB, and wall-clock render time.",
   "digest_skipped": null
  },
  {
   "id": "round-one-the-first-listen",
   "heading": "Round one — the first listen",
   "level": 2,
   "span": [
    5022,
    8075
   ],
   "chunks": [
    [
     5022,
     8075
    ]
   ],
   "chars": 3053,
   "digest": "a human operator rendered two requests on 2026-08-31 between 03:01 and 03:04 UTC. each request included three renders: nothing added, the adapter at full strength, and the adapter at 0.7. the untouched clip was preferred on both requests by the unblinded listener.",
   "digest_skipped": null
  },
  {
   "id": "round-two-the-caption-shape-probe",
   "heading": "Round two — the caption-shape probe",
   "level": 2,
   "span": [
    8075,
    11225
   ],
   "chunks": [
    [
     8075,
     11225
    ]
   ],
   "chars": 3150,
   "digest": "rendered on 2026-08-31 between 03:23 and 03:27 UTC, this probe tested two explanations using seven clips. it used the pass-five checkpoint and different adapter dials. the article reports that the loudness drop tracked the shape of the caption, not the adapter.",
   "digest_skipped": null
  },
  {
   "id": "round-three-the-one-artist-control",
   "heading": "Round three — the one-artist control",
   "level": 2,
   "span": [
    11225,
    13162
   ],
   "chunks": [
    [
     11225,
     13162
    ]
   ],
   "chars": 1937,
   "digest": "rendered on 2026-08-31 between 19:48 and 19:50 UTC, this control used a second adapter trained on one artist's ten releases. the listener found the non-adapter version better on Basement Loop 7, while the two were similar on Schwerelos (original mix).",
   "digest_skipped": null
  },
  {
   "id": "round-four-twelve-requests-four-ways",
   "heading": "Round four — twelve requests, four ways",
   "level": 2,
   "span": [
    13162,
    29751
   ],
   "chunks": [
    [
     13162,
     29751
    ]
   ],
   "chars": 16589,
   "digest": "this due-diligence round, rendered 2026-09-03 between 03:13 and 03:30 UTC, used twelve requests rendered four ways. each request was rendered with nothing added, the adapter at 0.5, the adapter at 1.0, and the one-artist adapter at 1.0.",
   "digest_skipped": null
  },
  {
   "id": "what-to-listen-for",
   "heading": "What to listen for",
   "level": 2,
   "span": [
    29751,
    31276
   ],
   "chunks": [
    [
     29751,
     31276
    ]
   ],
   "chars": 1525,
   "digest": "listeners can note if an adapter's clip has lost its floor, if loudness range doubles, or how the one-artist control sounds. eighteen of the sixty-five raw renders carry flat-topped stretches, and the one listen on record preferred the untouched clip in round one.",
   "digest_skipped": null
  },
  {
   "id": "how-to-check-these-clips",
   "heading": "How to check these clips",
   "level": 2,
   "span": [
    31276,
    33135
   ],
   "chunks": [
    [
     31276,
     33135
    ]
   ],
   "chars": 1859,
   "digest": "the audio/ directory contains raw renders and listening masters with corresponding manifests and fingerprints. the data kit includes request payloads and render records. of the sixty-six clips rendered, sixty-three files and players are served on this page.",
   "digest_skipped": null
  },
  {
   "id": "who-ran-this-and-thanks",
   "heading": "Who ran this, and thanks",
   "level": 2,
   "span": [
    33135,
    33664
   ],
   "chunks": [
    [
     33135,
     33664
    ]
   ],
   "chars": 529,
   "digest": null,
   "digest_skipped": "credits"
  }
 ],
 "chips": [
  {
   "id": "c-876817bf",
   "q": "how are the audio clips prepared for playback?",
   "a": "The players fetch a listening master for each clip. This master is the original raw render with one constant gain applied and nothing else, such as no limiter, compression, or equalization. The gain is determined by the smaller of the distance from −16 LUFS and the distance from a −1 dBTP ceiling.",
   "cites": [
    "how-to-listen"
   ],
   "published": false,
   "publish_state": "held",
   "quote": "",
   "quote_cite": ""
  },
  {
   "id": "c-b958c298",
   "q": "what was the outcome of the first listening round?",
   "a": "A human operator preferred the untouched clip on both requests during the first round. The 0.7 adapter renders were not ranked in this round, which serves as the basis for the article's verdict.",
   "cites": [
    "round-one-the-first-listen"
   ],
   "published": false,
   "publish_state": "held",
   "quote": "",
   "quote_cite": ""
  },
  {
   "id": "c-8b9d8fe7",
   "q": "what caused the loudness drop observed in round two?",
   "a": "The findings indicate that the loudness drop in round one tracked the shape of the caption rather than the adapter.",
   "cites": [
    "round-two-the-caption-shape-probe"
   ],
   "published": false,
   "publish_state": "held",
   "quote": "",
   "quote_cite": ""
  },
  {
   "id": "c-b3ac7145",
   "q": "how did the one-artist control compare to the original clips?",
   "a": "On Basement Loop 7, the non-adapter version sounded better because the adapter version added weird clippy hi hats. On Schwerelos (original mix), the two versions were considered pretty similar, though the listener eventually leaned slightly toward the adapter.",
   "cites": [
    "round-three-the-one-artist-control"
   ],
   "published": false,
   "publish_state": "held",
   "quote": "",
   "quote_cite": ""
  },
  {
   "id": "c-52299c11",
   "q": "how many requests were tested in the due-diligence round?",
   "a": "Twelve requests were written in advance and rendered four ways each: nothing added, the adapter at 0.5, the adapter at 1.0, and the one-artist adapter at 1.0.",
   "cites": [
    "round-four-twelve-requests-four-ways"
   ],
   "published": false,
   "publish_state": "held",
   "quote": "",
   "quote_cite": ""
  },
  {
   "id": "c-3aedbf66",
   "q": "what specific audio qualities should a listener notice?",
   "a": "Listeners may notice if an adapter's clip has lost its floor, whether the loudness range has doubled, or how the one-artist control sounds compared to the other models. Additionally, eighteen raw renders carry flat-topped stretches where samples are pinned at the peak.",
   "cites": [
    "what-to-listen-for"
   ],
   "published": false,
   "publish_state": "held",
   "quote": "",
   "quote_cite": ""
  },
  {
   "id": "c-ff17fb8b",
   "q": "how can the integrity of the audio files be verified?",
   "a": "The data kit contains a manifest with every file's full sha256 and a README with byte lengths. The listening masters have their own manifest and a JSON file containing the full recipe, including measured loudness, true peak, and the specific gain rules applied.",
   "cites": [
    "how-to-check-these-clips"
   ],
   "published": false,
   "publish_state": "held",
   "quote": "",
   "quote_cite": ""
  }
 ],
 "related": [
  {
   "slug": "hear-it-for-yourself",
   "why": "shares ground with § What to listen for · § How to listen"
  },
  {
   "slug": "ten-minutes-with-living-artists",
   "why": "shares ground with § What to listen for · § Twelve requests, four ways"
  },
  {
   "slug": "chopin-in-five-minutes",
   "why": "shares ground with § What to listen for · § The dial"
  }
 ],
 "thanks": "## Who ran this, and thanks {#who-ran-this-and-thanks}\n\nThe same small team and fleet of AI agents as the article, on the same machines, and the credits are [Ten Minutes with Living Artists](/ten-minutes-with-living-artists/)'s: the forty-one releases and their licences, the model, the trainer and the tools are all named there in full, deed by deed, because on this arc the credits are not a courtesy but the licence — [how that works](/how-we-work/) is next door, and the [licence ledger](/licences/) carries the model's row.",
 "kit": null,
 "seat_class": {
  "writer": "gemma-class",
  "judge": null,
  "writer_runtime": "vllm"
 },
 "bench": {
  "bullets_written": 3,
  "bullets_kept": 3,
  "digests_written": 7,
  "digests_kept": 7,
  "chips_written": 7,
  "chips_kept": 7,
  "chips_grounded": 0,
  "chips_published": 0,
  "dropped_by": {
   "new_noun": 0,
   "figure": 0,
   "length": 0,
   "cite": 0,
   "judge": 0,
   "quote": 0,
   "directive": 0,
   "redaction": 0,
   "profanity": 0
  },
  "new_noun_tokens_checked": [
   "0.5",
   "0.7",
   "01",
   "03",
   "04",
   "1",
   "1.0",
   "1.5's",
   "13",
   "13.29",
   "16",
   "19",
   "19.65",
   "2026-08-31",
   "2026-09-03",
   "23",
   "27",
   "30",
   "48",
   "50",
   "7",
   "ACE-Step",
   "Basement",
   "JSON",
   "LUFS",
   "Loop",
   "README",
   "Schwerelos",
   "UTC",
   "sha256"
  ],
  "new_noun_tokens_withheld": 0,
  "figure_definition": "v3",
  "drops": []
 },
 "generated_utc": "2026-09-09T01:25:35Z"
}