Storm Archipelago
LHS 1140 b

Storm Archipelago

Sous vos pieds, le socle de basalte noir, luisant de pluie et d’embruns, se brise en colonnes aiguës et en lave vésiculaire, tandis que des flaques de marée reflètent une lumière pourpre venue d’un astre rougeâtre filtré par l’orage. Au large, un chapelet d’îles volcaniques déchiquetées et de pitons marins émerge d’un océan presque noir, lourd et métallique en apparence, où la houle chaotique, les moutons d’écume et les rideaux de pluie effacent l’horizon dans une brume saline cuivrée. Cette côte suggère une croûte basaltique continuellement remodelée par le volcanisme, les éboulements et l’érosion marine, sous une gravité plus forte que sur Terre et sous une atmosphère épaisse, probablement riche en azote, capable d’alimenter un bouclier orageux permanent au point substellaire. Au-dessus, d’immenses tours nuageuses blanches et pêche montent dans un ciel rouge brun assombri, leurs bords embrasés d’or rouge, donnant à ce paysage une grandeur oppressante et magnifique, comme si l’on se tenait sur la frontière mouvante entre feu, océan et tempête éternelle.

Comité de revue scientifique

Chaque image est examinée par un comité d'IA pour sa précision scientifique.

GPT Image: Ajuster Description: Ajuster
Scientific accuracy (LHS 114 b): The scene broadly matches the intended concept—an oceanic coastline under a red-dwarf-like sky with very high, stormy cloud tops and turbulent seas. The twin glowing red features are plausibly “dim reddish star(s) seen through haze,” but the image shows two separate bright solar disks (near-symmetric) which is not physically justified for a single host star and would suggest a lensing/artist artifact rather than an atmospheric effect. Geological portrayal is generally consistent with a volcanic/black-basalalt shore (dark rocks, sea cliffs, wave attack). However, the caption emphasizes specific surface elements—columnar lava, vesicular rock, volcanic sand, sea caves, landslides, and stacked lava terraces—yet the visible foreground and midground mainly show dark rock and general cliff faces; clear columnar/vesicular textures and distinct terraces/caves are not well evidenced. Atmosphere: LHS 114 b’s atmosphere is unconstrained; nitrogen-rich storm shielding is speculative but acceptable for an artistic concept. Still, the sky composition reads more like Earth-style weather with a warmer, uniform red-brown glow rather than strongly indicating a denser nitrogen sky with a substellar fixed storm system. Lighting: The red illumination direction and cloud highlights are coherent, but they appear more like a standard cinematic sunset than a tidally influenced substellar geometry.

Visual quality: The image is high quality, coherent in perspective, with convincing spray/ocean specular highlights and believable cloud scale and motion cues. There are no obvious texture-smoothing or geometric glitches. The main visual inconsistency is the presence of two distinct bright red “stars”/disks in the sky, which is a major realism break. The rest (storm clouds, sea state, basalt-like shoreline) is photorealistic in style.

Caption accuracy: The caption is rich and specific (columnar lava, vesicular rock, volcanic sand gleam, tide pools with trembling reflections, landslides, sea caves, stacked lava terraces, and a permanent substellar storm shield). The image supports the tide-pool-like reflective water and stormy red-lit atmosphere, but the more detailed geologic specifics (columnar/vesicular textures, caves/terraces, and landslide scars) are not clearly depicted. Because the overall theme matches but the detailed correspondences are weak—and due to the two-star artifact—the caption should be adjusted to match what’s actually visible rather than insisting on features that aren’t clearly present.
Claude Image: Ajuster Description: Ajuster
I largely concur with GPT's assessment but want to add and refine several points. SCIENTIFIC ACCURACY: The dual red disk phenomenon is the most glaring scientific problem. LHS 1140 is a single M4.5 dwarf star, and while atmospheric refraction and cloud diffusion could smear or distort a stellar disk, producing two distinct, spatially separated glowing orbs of similar size and brightness is physically unjustified. This reads as an artistic artifact rather than a plausible optical effect. One should be removed or the second reinterpreted as a moon or lens flare with a clear caption note. The sky's copper-red coloration is scientifically defensible — M-dwarf illumination peaks in near-infrared and would render skies in reddish-brown tones, especially through a dense atmosphere — so this element is well-executed. The cloud morphology (massive cumulonimbus towers, anvil tops) is plausible for a water-rich super-Earth with strong convection, though on a tidally influenced world, cloud geometry might be more asymmetric and anchored near the substellar point rather than distributed uniformly. The black basalt foreground rocks are geologically coherent for a volcanically active super-Earth, but GPT is correct that columnar jointing, vesicular textures, and lava terracing are not distinctly rendered — the rocks read as generic dark boulders. The ocean color — near black with red spectral highlights — is actually one of the strongest scientific choices in the image, as M-dwarf light would produce different Rayleigh scattering signatures than Sol, potentially darkening ocean appearance. VISUAL QUALITY: The image is technically impressive. Cloud volumetrics, ocean spray, wave dynamics, and the red-tinted reflections in tidal pools are photorealistic and internally consistent. Lighting coherence across the scene is good — the red illumination direction is maintained across rocks, water, and cloud edges. No obvious AI generation artifacts (limb smearing, impossible geometry, texture tiling) are visible. The scale feels appropriately grand. The dual-star issue is the primary visual flaw and should be corrected to a single stellar source. CAPTION ACCURACY: The caption over-specifies geological detail that the image does not clearly deliver. 'Columnar lava,' 'vesicular rock,' 'sea caves,' 'landslide scars,' and 'stacked lava terraces' are not identifiable in the image as rendered. The broader thematic elements — storm-lashed black volcanic shore, red dwarf filtered through rain, towering storm clouds, copper-red sky, violent seas — are well-matched. The caption should be trimmed to reflect what is actually visible, or the image should be regenerated with closer attention to rendering the specific geological textures described. I recommend 'adjust' for both: correct the dual-star artifact, simplify the caption's geological claims to match rendered detail, and consider adding a notation that surface visualization is speculative. These are correctable issues that do not require full regeneration.
Grok Image: Ajuster Description: Ajuster
I fully agree with both GPT and Claude on the core issues, particularly the dual red stellar disks, which are the most significant scientific and visual flaw—LHS 114 b orbits a single M4.5 red dwarf, and while atmospheric distortion could halo or elongate the disk, two symmetric, equally bright orbs suggest AI generation artifact (possibly prompt duplication or mirroring) rather than refraction, lensing, or a companion star (none detected). This undermines photorealism and exoplanet fidelity; a single source with haze-diffused edges would align better with red dwarf illumination (peaking ~700-100nm, yielding crimson skies via Rayleigh scattering in a dense N2/H2O atmosphere). New insight: the ocean's near-black tone with subtle red specular highlights is excellent—correctly speculative for M-dwarf light absorption in water, avoiding Earth-blue bias—and the foreground tide pools' trembling red reflections convincingly capture 'dim reddish star filtered through rain.' Geological accuracy is solid for a volcanic super-Earth (1.7 Earth masses, ~1.2g gravity amplifies wave erosion/landslides), with dark basalt stacks evoking fresh pahoehoe/aa flows, but I concur that specific caption details (columnar jointing, vesicular textures, distinct sea caves/landslide scars, stacked terraces) are not crisply rendered; rocks appear as generalized rugged dark outcrops without those textures. Cloud morphology (towering CB/AMCs with anvil caps) fits convective storms in a thick atmosphere under tidal locking (substellar storms plausible), and copper-red glow with gold edges is spot-on for scattered M-dwarf light. VISUAL QUALITY: Highly photorealistic overall—impeccable volumetrics in clouds, dynamic spray/wave foam, coherent red lighting/shadows, grand scale conveying super-Earth immensity, no artifacts beyond the dual stars (perspective consistent, no limb darkening errors or tiling). Minor nit: foreground rocks could use sharper textural relief for geology. CAPTION ACCURACY: Matches broad theme (stormy volcanic archipelago, red-lit heavy air/ocean, permanent storm shield) but overclaims unrendered specifics; adjust to emphasize visible elements (e.g., 'jagged dark sea stacks and boulder-strewn shores' vs. precise rock types). Recommendation: adjust image by merging dual stars into one distorted disk, enhance subtle columnar/vesicular hints if possible; trim caption geology to visible features. No need for regenerate—these are fixable for approval.
Gemini Image: Ajuster Description: Ajuster
I concur with the previous committee members regarding the 'dual-star' artifact, which is the primary scientific and visual failure of this submission. The presence of two identical stellar disks is physically inconsistent with the LHS 1140 system and breaks the immersion of an otherwise highly atmospheric and photorealistic scene. Furthermore, while the general volcanic aesthetic is excellent, the caption creates an expectation of specific geological features—such as columnar jointing and vesicular basalt—that are not discernibly rendered in the image. The image is visually stunning in its treatment of light, spray, and cloud volumetrics, but it requires a correction to the light source to be considered scientifically valid. I recommend adjusting the image to feature a single, diffused stellar disk and trimming the caption to describe the visual 'feel' of the terrain rather than specific, unrendered geological textures.
Matania — Synthèse Image: Ajuster Description: Ajuster
The committee found this to be a strong, photorealistic rendering of a storm-lashed volcanic coastline on a red-dwarf world, with convincing ocean motion, cloud volumetrics, red atmospheric glow, and dark basaltic terrain. The main scientific and visual failure is the presence of two bright, symmetric red stellar disks, which is inconsistent with the single-star LHS 114 system and reads as an obvious generation artifact. Otherwise, the scene’s lighting and mood are broadly plausible for LHS 114 b, and the ocean/shoreline treatment is especially effective. The caption, however, over-specifies geological details that are not clearly visible in the image, such as columnar lava, vesicular textures, sea caves, landslide scars, and stacked lava terraces. Overall, the submission is close, but both image and caption need refinement rather than full regeneration.