One sentence of art direction becomes a cohesive eight-shot campaign — then a packshot, a motion teaser, and a finished reel. Built on Krea 2 Open-Source, the most aesthetic open image model on the network, and the first open base you can train.
Every shot shares one byte-identical art-direction prefix — “ivory, sand, warm stone & oat palette; soft diffused north-window light; sculptural wool & linen; quiet luxury; fine film grain; Vogue-cover restraint; medium-format.” Only the trailing subject changes. The palette, light and grain hold across all eight.








All eight rendered on sdk.daydream.monster via krea-2-os (fal-ai/krea-2/turbo) — ~3 seconds per shot, ~$0.25 for the set.
The packshot above is a one-call bg-remove on the accessory still — a clean transparent cutout for e-commerce. The teaser animates the hero portrait; the reel stitches the campaign into a platform cut. Add sonilo-t2m for a licensed soundtrack and ffmpeg-export to fan out 9:16 / 1:1 / 16:9.
Eight shots from one sentence already feel like a campaign. To make it permanent — so every future look is on-brand by construction across 8, 80, or 800 shots — freeze the aesthetic into a LoRA. Krea 2 Open-Source is the first open base on the network you can train, and the one to pick when look matters more than a photoreal identity.
krea-2-os). The set above.submit_lora_train · base_model:"krea-2" → trains on fal-ai/krea-2-trainer. ~6–8 min, ~$22, once.create_media · model_override:"krea-2-lora" — every new look inherits the house aesthetic.# Train your brand aesthetic into a reusable LoRA — across every surface
# MCP
submit_lora_train({ zip_url, trigger_word:"ATELIERNORD", base_model:"krea-2", steps:500 })
apply_lora({ weights_url, trigger:"ATELIERNORD", prompt:"a new SS26 look", model:"krea-2-lora" })
# CLI
livepeer lora train --zip <url> --trigger ATELIERNORD --base krea-2 --steps 500
livepeer lora apply <weights_url> --trigger ATELIERNORD --prompt "a new SS26 look" --model krea-2-lora
# Webapp: right-click canvas → Train LoRA → base = "Krea 2 (Open-Source)" → /lora apply <name>
Honest status (2026-06-24): the krea-2-lora-training cap is registered, accepts jobs and meters them; full weight-retrieval through the SDK currently hits a known 202-polling follow-up (the orchestrator returns HTTP 202 for async training and the SDK's status poller doesn't yet treat 202 as “still running”). The recipe above is wired across MCP + CLI + webapp and lights up the moment that SDK fix lands. The campaign and finishing passes shown on this page are fully proven, end-to-end.
The aesthetic sentence is the whole product. Swap it and the same pipeline ships a different vertical:
Why it matters. Aesthetic used to be the moat closed apps charged for. Krea 2 Open-Source puts the most tasteful open image base on a metered network — and makes it trainable, so a small brand can lock its look once and out-produce a studio. The campaign on this page cost about a quarter and ran in under a minute.
Follow it yourself. Open the Train Your Brand Aesthetic playbook, fill in one aesthetic sentence, and run it.