card: consolidate img2img into a single table

README.md

@@ -40,84 +40,56 @@ Full single-file SDXL checkpoints, ready to load as-is.
 
 
 
+
 ---
 
 ## img2img
 
-Both checkpoints do `img2img` as well as `txt2img`. Same input, same prompt, same seed, through each.
+Both checkpoints do `img2img` as well as `txt2img`. Two sources, each through both checkpoints.
 
 <table>
-<tr><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/img2img-source.png" width="250" alt="source prompt: a red apple on a rustic wooden table beside a window, soft daylight, visible wood grain, studio photograph, sharp focus — 1024×1024, seed 7 — generated separately as the img2img input"></td><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/img2img-juggernaut.png" width="250" alt="img2img prompt: a still life oil painting of a red apple on a rustic wooden table beside a sunlit window, thick visible oil paint brushstrokes, warm daylight, painterly canvas texture, high detail — Juggernaut XL v9, 1024×1024, 25 steps, guidance 7.5, strength 0.75, seed 42 — about 174 s on an A100 80GB"></td><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/img2img-illustrious.png" width="250" alt="img2img prompt: a still life oil painting of a red apple on a rustic wooden table beside a sunlit window, thick visible oil paint brushstrokes, warm daylight, painterly canvas texture, high detail — Illustrious Anime v4, 1024×1024, 25 steps, guidance 7.5, strength 0.75, seed 42 — about 174 s on an A100 80GB"></td></tr>
-<tr><td><sub>**source prompt:** `a red apple on a rustic wooden table beside a window, soft daylight, visible wood grain, studio photograph, sharp focus` — 1024×1024, seed 7 — generated separately as the img2img input</sub></td><td><sub>**img2img prompt:** `a still life oil painting of a red apple on a rustic wooden table beside a sunlit window, thick visible oil paint brushstrokes, warm daylight, painterly canvas texture, high detail` — **Juggernaut XL v9**, 1024×1024, **25 steps**, **guidance 7.5**, **strength 0.75**, seed 42 — about 174 s on an A100 80GB</sub></td><td><sub>**img2img prompt:** `a still life oil painting of a red apple on a rustic wooden table beside a sunlit window, thick visible oil paint brushstrokes, warm daylight, painterly canvas texture, high detail` — **Illustrious Anime v4**, 1024×1024, **25 steps**, **guidance 7.5**, **strength 0.75**, seed 42 — about 174 s on an A100 80GB</sub></td></tr>
+<tr><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/img2img-source.png" width="250" alt="photographic source"></td><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/img2img-juggernaut.png" width="250" alt="Juggernaut XL v9 on the photographic source"></td><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/img2img-illustrious.png" width="250" alt="Illustrious Anime v4 on the photographic source"></td></tr>
+<tr><td><sub>**source prompt:** `a red apple on a rustic wooden table beside a window, soft daylight, visible wood grain, studio photograph, sharp focus` — 1024×1024, seed 7</sub></td><td><sub>**img2img prompt:** `a still life oil painting of a red apple on a rustic wooden table beside a sunlit window, thick visible oil paint brushstrokes, warm daylight, painterly canvas texture, high detail` — **Juggernaut XL v9**, 1024×1024, **30 steps**, **strength 0.75**, seed 7 — about 170 s on an A100 80GB</sub></td><td><sub>**img2img prompt:** `a still life oil painting of a red apple on a rustic wooden table beside a sunlit window, thick visible oil paint brushstrokes, warm daylight, painterly canvas texture, high detail` — **Illustrious Anime v4**, 1024×1024, **30 steps**, **strength 0.75**, seed 7 — about 170 s on an A100 80GB</sub></td></tr>
+<tr><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/fixture-source.png" width="250" alt="tests/sdxl/test_06_source.png, a synthetic gradient"></td><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/fixture-juggernaut.png" width="250" alt="Juggernaut on the test_06 fixture"></td><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/fixture-illustrious.png" width="250" alt="Illustrious on the test_06 fixture"></td></tr>
+<tr><td><sub>**source** — `tests/sdxl/test_06_source.png`, a synthetic gradient with a flat circle</sub></td><td><sub>**img2img prompt:** `an oil painting of an abstract composition with a large glowing golden circle at the centre over a flowing rainbow gradient, thick visible brushstrokes, impasto ridges catching the light, visible canvas texture` — **Juggernaut XL v9**, 1024×1024, **25 steps**, **guidance 7.5**, **strength 0.75**, seed 42 — about 174 s</sub></td><td><sub>**img2img prompt:** `an oil painting of an abstract composition with a large glowing golden circle at the centre over a flowing rainbow gradient, thick visible brushstrokes, impasto ridges catching the light, visible canvas texture` — **Illustrious Anime v4**, 1024×1024, **25 steps**, **guidance 7.5**, **strength 0.75**, seed 42 — about 174 s</sub></td></tr>
 </table>
 
 ### Describe the image you want. Do not give an instruction.
 
-This is the single thing most likely to waste your afternoon, so it is first.
-
 SDXL has no instruction-following training. The prompt is a **description of the
-image to produce**, not a command to apply to the source. Write what the finished
-picture looks like:
-
-> ✅ `a still life oil painting of a red apple on a rustic wooden table beside a sunlit window, thick visible oil paint brushstrokes…`
-
-not
+image to produce**, not a command applied to the source:
 
-> ❌ `the same scene rendered as a vivid oil painting, thick visible brushstrokes…`
+> ✅ `a still life oil painting of a red apple on a rustic wooden table beside a sunlit window…`
+>
+> ❌ `the same scene rendered as a vivid oil painting…`
 
-The instruction form is taken literally and produces exactly what it says. Asking
-for "the same scene rendered as an oil painting" returned an apple **coated in
-dripping paint** with a paintbrush in frame, and at higher strength a different
-scene altogether. Nothing was wrong with the model, the checkpoint or the
-settings — only the prompt. The samples above use the descriptive form at the
-same `strength 0.75` that produced the broken results with an instruction.
+The instruction form is taken literally. Asking for "the same scene rendered as an
+oil painting" returned an apple coated in dripping paint with a paintbrush in frame,
+and at higher strength a different scene altogether — at the same `strength 0.75`
+that produced the top row here.
 
 ### The two checkpoints
 
-Both keep the subject at 0.75 — same apple, same position, same scale, same light.
-They differ in treatment:
-
-- **Juggernaut XL v9** — heavier impasto, warmer, more overtly painted. The
-  photographic checkpoint, and the one to reach for on photographic sources.
-- **Illustrious Anime v4** — cleaner and brighter (it retains ~93% of the source's
-  luminance against Juggernaut's ~85%), rendered more as illustration than as oil
-  on canvas. Being an anime checkpoint, it leans that way even when asked for a
-  painting.
-
-### On `strength`
-
-It is the fraction of the image the sampler may rewrite, trading style against
-fidelity. Swept from 15 to 40 steps and guidance 5.0 to 12.5 at `strength 0.75`,
-output was stable across the whole range — so with a descriptive prompt these
-settings are forgiving. Below ~0.3 you get little more than a gloss pass; above
-~0.85 the source becomes a loose suggestion.
+Both keep the subject at 0.75 — same apple, position, scale and light. Juggernaut
+is heavier impasto and warmer; Illustrious is cleaner and brighter (retaining ~93%
+of source luminance against Juggernaut's ~85%), rendering more as illustration.
 
+### `strength` and the swept range
 
+Swept 15–40 steps and guidance 5.0–12.5 at `strength 0.75`, output was stable
+throughout. Below ~0.3 gives little more than a gloss pass; above ~0.85 the source
+becomes a loose suggestion.
 
+### On the synthetic source
 
-### Test fixture
-
-`tests/sdxl/test_06_source.png` through both checkpoints.
-
-<table>
-<tr><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/fixture-source.png" width="250" alt="test_06 fixture source, a synthetic gradient with a flat circle"></td><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/fixture-juggernaut.png" width="250" alt="Juggernaut XL v9 on the test_06 fixture source"></td><td><img src="https://huggingface.co/ChrisColeTech/sdxl/resolve/main/samples/fixture-illustrious.png" width="250" alt="Illustrious Anime v4 on the test_06 fixture source"></td></tr>
-<tr><td><sub>**source** — `tests/sdxl/test_06_source.png`, a synthetic gradient with a flat circle</sub></td><td><sub>**img2img prompt:** `an oil painting of an abstract composition with a large glowing golden circle at the centre over a flowing rainbow gradient, thick visible brushstrokes, impasto ridges catching the light, visible canvas texture` — **Juggernaut XL v9**, 1024×1024, **25 steps**, **guidance 7.5**, **strength 0.75**, seed 42 — about 174 s on an A100 80GB</sub></td><td><sub>**img2img prompt:** `an oil painting of an abstract composition with a large glowing golden circle at the centre over a flowing rainbow gradient, thick visible brushstrokes, impasto ridges catching the light, visible canvas texture` — **Illustrious Anime v4**, 1024×1024, **25 steps**, **guidance 7.5**, **strength 0.75**, seed 42 — about 174 s on an A100 80GB</sub></td></tr>
-</table>
-
-Measured against the source: Juggernaut 1.00× luminance, Illustrious 1.03×.
-Juggernaut applies vertical brushwork across the gradient; Illustrious leaves it
-close to the source. In both, the flat circle is reproduced flat — a uniform
-region carries no detail for the sampler to work from at `strength 0.75`, while
-the graduated area does.
-
-Two properties of this source affect how it should be used in testing:
-
-- With no subject present, subject loss cannot be observed. `strength 0.90` passes
-  against this fixture and does not preserve the subject of a photograph.
-- `histogram_correlation` measures 0.001 and −0.017 here, against 0.73–0.78 for
-  the same runs on a photograph, because a near-uniform histogram has little to
-  correlate against. `TEST_PLAN.md` treats a negative value as a VAE scaling
-  symptom.
+Bottom row: Juggernaut 1.00× source luminance, Illustrious 1.03×. Juggernaut
+applies vertical brushwork to the gradient; Illustrious leaves it close to the
+source. In both, the flat circle is reproduced flat — a uniform region carries no
+detail for the sampler at `strength 0.75`, while the graduated area does. Two
+consequences for testing: subject loss cannot be observed against a subjectless
+source (`strength 0.90` passes here and does not preserve a photograph's subject),
+and `histogram_correlation` measures 0.001/−0.017 here against 0.73–0.78 for the
+same runs on a photograph.
 
 ## Recommended settings