night train

ENGLISH

I created n1ght tr41n as a background LoRA concept for a fictional European luxury sleeper express of the late 1920s and early 1930s. I wanted the romance of a train racing through the night: polished walnut, brass, crimson upholstery, soft lamps and moonlit landscapes sliding past the windows.

I built the set around eleven distinct locations: the bar car, restaurant car, deluxe sleeper cabin, side corridor, baggage car, kitchen, smoking salon, piano car, open rear platform, panorama observation car and steam locomotive footplate. Each location has four compositions, with its own floor, lighting and atmosphere. A final aerial view brings the whole train together on a moonlit stone viaduct.

These are empty backgrounds, without people. The train is generic and fictional; I did not base its identity on a real railway service or brand. The set contains 45 square images and matching captions.

Trigger: n1ght tr41n

I place the exact trigger at the beginning of the prompt, then describe the carriage, camera angle, materials and night lighting.

Example prompt structures:

n1ght tr41n, luxury sleeper express bar car, hammered brass counter, walnut marquetry, crimson runner over parquet, warm opaline lamps, blurred moving night landscape, cinematic photography

n1ght tr41n, deluxe sleeper cabin, ivory bedding, brass luggage rack, patterned burgundy carpet, shaded bedside lamp, moonlit fields outside the window, low angle

n1ght tr41n, steam locomotive footplate, brass gauges, steel levers, glowing firebox, dark ribbed steel floor, cinematic night lighting

Suggested starting points only, not verified results:

For a future SDXL version, I would start testing around 25-35 steps, CFG 4-6 and LoRA weight 0.6-0.9, then adjust to the actual checkpoint and sampler. For a separately trained Illustrious version, these would also be exploratory values, subject to that checkpoint's own guidance. The finished weights, compatibility and settings have not been tested; this deliverable is the dataset, not a trained model.

For Krea, I cannot confirm that this future LoRA can be loaded or that steps, CFG or weight controls will be available. I would first verify support for the trained base model and custom LoRA in the selected workflow. The scene prompts can serve as visual starting points, without implying confirmed trigger or LoRA support.

arrssenne

FRANCAIS

J'ai cree n1ght tr41n comme concept de LoRA de backgrounds pour un train de nuit europeen fictif de luxe, inspire de la fin des annees 1920 et du debut des annees 1930. Je voulais retrouver le romantisme d'un express filant dans la nuit : noyer poli, laiton, tissus cramoisis, lampes douces et paysages au clair de lune qui defilent aux fenetres.

J'ai organise le set autour de onze lieux distincts : wagon-bar, restaurant, cabine deluxe, corridor lateral, wagon a bagages, cuisine, salon fumeur, wagon-piano, plateforme arriere ouverte, wagon panoramique et cabine de locomotive a vapeur. Chaque lieu propose quatre compositions avec son propre sol, son eclairage et son ambiance. Une vue aerienne finale rassemble le train entier sur un viaduc de pierre au clair de lune.

Ce sont des backgrounds vides, sans personnes. Le train est generique et fictif; son identite ne reprend aucun service ferroviaire reel ni aucune marque. Le set contient 45 images carrees avec leurs captions.

Trigger : n1ght tr41n

Je place le trigger exact au debut du prompt, puis je decris le wagon, le cadrage, les materiaux et la lumiere nocturne. Les exemples anglais ci-dessus montrent cette structure.

Points de depart suggeres seulement, sans validation :

Pour une future version SDXL, je commencerais les essais autour de 25-35 steps, CFG 4-6 et un poids LoRA de 0.6-0.9, a ajuster selon le checkpoint et le sampler. Pour une version Illustrious entrainee separement, ces valeurs resteraient exploratoires et devraient suivre les recommandations du checkpoint utilise. Aucun poids final, resultat, reglage ni compatibilite n'a ete teste; cette livraison est un dataset, pas un modele entraine.

Pour Krea, je ne peux pas confirmer le chargement de ce futur LoRA ni la disponibilite des controles steps, CFG ou poids. Je verifierais d'abord la prise en charge du modele de base et du LoRA personnalise dans le workflow choisi. Les prompts de scene peuvent servir de points de depart visuels, sans promettre la prise en charge du trigger ou du LoRA.

arrssenne