Provenance · verbatim
Every prompt, in full.
Vespera is built on a single generated film. Here is the exact prompt and settings used to make it, published for provenance.
The film
Slow, steady forward dolly gliding through an otherworldly twilight garden at dusk. Enormous luminous night-blooming flowers unfurl in slow motion, glowing amber and coral pollen drifting through the warm hazy air. Deep plum and oxblood sky above, soft jade foliage below, volumetric golden light rays cutting through mist, embers floating gently upward. Dreamy, surreal, painterly atmosphere. The camera creeps forward at one steady, almost imperceptible pace, no hard cuts, one single continuous motion, calm and hypnotic. Cinematic 24fps film look, ultra detailed, shallow depth of field. Seamless loop, first and last frame matching, position and lighting unchanged.
Settings and provenance
- Model: Kling 3.0 (Higgsfield), text to video, standard mode.
- Output: 1280 by 720, 16:9, 10 seconds, silent.
- Grounding rules: one tempo, subject plus movement stated explicitly, seamless loop closing line, so the scrub reads as one continuous drift.
- Post: sliced into 90 WebP frames per screen size with ffmpeg; no other imagery is generated.
- Disclosure: under EU AI Act article 50, generated media is disclosed in the footer of every page.
Everything else on this site, the type, the layout, the scrub engine, is written by hand. Read the build guide to reproduce it.