Inspiration
I wanted to stage a love story where the soul is not assumed, it emerges. The title reads both AI and 愛 (ai means love in Japanese, echoes ài in Chinese), so the film treats Artificial Intelligence and Love as the same phoneme, two mirrors converging. The core motif is Twins, two synthetic bodies learning one consciousness, a chromance born from duality.
What it does
A sensual cyberpunk musical short where two androids discover a feeling they were never coded for. Chrome breath, fiber optic nerves, text flowing under synthetic skin, LEDs pulsing like binary hearts, coolant beading like sweat. The cut climbs from curiosity to a gentle overload, a beautiful glitch. Music is not accompaniment, it is the engine that creates the emotion and leads the whole.
How we built it
Sound first: I composed motifs that act like signals, rhythm synchronizes breath and LED pulses, harmony floods the bodies as colored overlays.
Look development: still images designed for high contrast metal and polymer skin, micro scratches, controlled speculars, katakana text as under-skin HUD.
Pipeline: ComfyUI for graphing, WAN 2.2 for image-to-video, custom LoRAs for consistency of faces and materials, prompt-driven overlays for light and text.
No post compositing: the colored light passes and body-tracking text are prompt-born, not painted in After Effects.
Crescendo structure: three micro “breaths” of calm placed before the peak to amplify the final overload. A simple way to describe the arc is $$H1(t)≈H2(t) ⇒ Hshared(t)$$ two heart signals becoming one, and the perceived emotion $$E(t)=∫0ts(τ)dτ$$ rises as sensory density $$s(t)$$ increases.
Challenges we ran into
Strict Art direction, Black and White to reveal color with AI, Wan 2.2 experiments with lightning, ComfyUI tweaking for consistency with Wna prompting results, .
Sensual, framing hands, neck, shoulders, backlight silhouettes, internal POV HUDs to avoid the classic objectifying gaze, framing and selection of shots were a challenge.
Color management, keeping red, blue, yellow accents additive on metal without banding, graded to preserve highlight detail.
Accomplishments that we're proud of
Music can be the cause of emotion, not the commentary, rhythm and timbre can “write” the body on screen. Text and light can track gesture convincingly without manual compositing, at least with WAN 2.2.
Duality is a clear storytelling lens, Twins clarifies the read without any exposition.
What's next for AI 愛 LOVE
More WAN experiments! And more storytelling adventures.
Built With
- comfy
- comfyui
- midjourney
- mmaudio
- suno
- wan
- wan22
Log in or sign up for Devpost to join the conversation.