FIND YOUR SELF-ISH

2024
Short Film

<intro>



"With a mirror on the wall, this is how we fall.
Find your self, be your selfish.
Slices of life, cascade down the everlasting staircase, only to be persued relentlessly by their very own reflections.
There is life outside, or is it?
Find your self,be your selfish."

[Film quote]
   ↓ Movie Poster "Find Your Self-Ish":
Movie Poster for "Find Your Self-Ish". An avant-garde short film, merging CGI and AI into an eerie journey of self-reflection. Very experimental and abstract visual style, utilizing frame in-consistency of stable diffusion. It feels very eery and nightmarish.

↓ Short video scenes out of "Find Your Self-Ish" (2024):

↓ Still Frames of "Find Your Self-Ish" (2024):

Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185
Redshift
185A
185
185

/ / Project Context



An avant-garde short film, merging CGI and AI into an eerie journey of self-reflection.

FIND YOUR SELF-ISH is an autobiographical and experimental CG film shot in first person. It explores themes of internal struggle and distress. The film is based on the filmmaker's own experiences with anxiety and addiction. The director experiments with cutting-edge AI techniques, to generate visuals that are being morphed with traditional 3D animation,  enhancing the visual storytelling uniquely. The abstract narrative of the movie is driven by the sensations of the protagonist, which manifest themselves in visual distortions and an evolving environment. Without dialogue, there is a big emphasis on the visual narrative and intra-diegetic soundscape.

The movie embraces the aesthetics of generative AI, to distort the image in innovative ways and combines this technique with other experimental manipulation methods. This makes the lived experience of the protagonist tangible for the audience. The abstract storytelling and first-person perspective invite the audience to project their own experiences onto the narrative, which opens the movie to a wide range of interpretations. Even the characters are non-specific in their appearance and morph from one AI-generated person into another. The stunning visuals are complemented by an atmospheric synth soundtrack.


//Project Info

data = {
    "date": "2024-03"
    "type": "personal"
    "contributor": "Benno Schulze"
    
    "category": [
        "EXPERIMENTAL",
        "SHORT_FILM",
        "AVANT_GARDE"
        "STABLE_DIFFUSION"
   ]
}

<INSIGHT>

Note: This section consists of concept art, testrenders, technical aspects, previz animations and other work in progress material,showcasing the steady improvement of quality and general workflow behind the scenes.

/ Styleframing +
Concept Art

Concept render for an avant-garde short film, merging CGI and AI into an eerie journey of self-reflection. Very experimental and abstract visual style. The concept render shows a burried body, although the color and mood is positive and calming. The visual composition with the burried face looks kind of eery and scary.
[SCROLL]
[SLIDE] > > >
↑  Visualizing the thematic ideas and overall stylistic vision I have in mind for the movie.
↑  Test on visualizing feeling of distress / panic
↑  Concept on implementing Color / Projections as a means of Visual Storytelling (Inspired by the drug (heroin) scene in Trainspotting 2)
↑  Digital Styleframing, especially for color composition
Concept render for an avant-garde short film, merging CGI and AI into an eerie journey of self-reflection. Very experimental and abstract visual style. The Picture looks moody and calming, 2 people lying in bed, their feet sticking out.
↑  Digital Styleframing, testing a technical approach to morph CGI with Stable Diffusion, very basic and early stage with a simple overlay.
↑  Testing poses using adjusted poses and human model from Mixamo, edited using a Volume Builder + Mesher.
Custom Material utilizing various shaders (like Curvature + Ramp) and layering.
↑  Testing poses using adjusted poses and human model from Mixamo, edited using a Volume Builder + Mesher.
Custom Material utilizing various shaders (like Curvature + Ramp) and layering.

Styleframing, inspired by an exhibition by Phyllia Barlow in Hannover, the last one before her death.
↑  Styleframing on world building / architectual style.
↑  Styleframing on world building / architectual style.
↓  Experimenting with rough textures and shapes on paper (Acrylic Paint + Charcoal Pencil on Paper):
(left) → 
Experimenting on combining Acrylic Paint and String on Paper, inspired by Anselm (2023, Wenders, Wim)
(right) → 
Working on Sandstone using Hammer and Chisel, further experimenting on compositions and shapes.

/ / Testing with Stable Diffusion

(left)
First testing with Stable Diffusion + Control Net, utilizing its facetracking features (input by me, overlay calculated by SD)
(right)
Output by SD, with my facetrack given as an input.
(left) →  
Test on a custom checkpoint, trained on pictures of myself. The results are not as desired to say the least... (All trained and running locally, no web connection)
(right) →
Second test on a custom checkpoint, trained on pictures of myself. Input dataset of roughly 100 pictures of various backgrounds and poses. (All trained and running locally, no web connection)
↓  Test on Deforum + StableDiffusion
↓  Second test on using SD + ControlNet body + facetracking, this time on a animation
↓  Experimental test on giving SD very few sampling steps, resulting in rough shapes.
(left)
Segmentation input sequence (rendered in C4D)
(right)
Resulting output sequence processed by SD

Experimenting on the idea of shifting environments, buildings scaling in size, people disappearing. In the end: left alone on a street through nothing. Segmentation input was rendered in C4D.
  Testing changing atmosphere (daytime switch)

/ / Technical Foundation (Cinema4D + Redshift)

↓  Camera Mapping SD Output inside Redshift as a means of stress inducing visual narrative (Test)
Sketch and segmentation maps rendered with a separate camera, which is locked to the moving POV camera of the final render. Rotation and zoom are manually keyframed to ensure the head and body consistently fill the frame.
↓  Output generated by Stable Diffusion. The person depicted is not meant to be analyzed by appearance; rather, their shifting gender, ethnicity, and other traits serve to convey a generalized, symbolic portrayal of people as a whole. This was achieved using the Dynamic Prompt Addon inside Stable DIffusion
↓  Alpha matte for keying out the background
↓  Camera mapped SD output projected back onto the original person.
↓  Enabled Subsurface Scattering in order to create a semi realistic lighting behaviour fitting to the 3D environment
↓  Early Character Rig, Camera Rigged to Head Movement:
↓  Fake cloud shadows using combination of animated noise-maps and Volume Builder
↓  Faked Cloud Shadow using simple layered and animated noise maps
↓  Volume Cloud visible to the camera
↓ Test to morph SD Output inside Redshift (Camera Mapping)
↓ Procedural cigarette bud, which can be keyframed to a certain "smoke up" threshold using a combination of Emitters, Volume Builder / Mesher and Boolean.
↓ Custom made Skin Shader and Hair Rendering with realistic Subdivision Surface Scattering
Testing limits of Hair Render, not be seen in the film

<World Building>

Note: This section contains raw renders with camera angles and bright daylight that are not part of the final short film. As a result, some parts may appear lower in quality due to being optimized for distant camera viewpoints (LOD) and/or depth of field effects.

Find Your Self-Ish maintains a continuous shot throughout the entire short film, yet the narrative embeds a hard but seamless shift in the scenery. While certain aspects of the environment and overall layout remain consistent, there is a stark transformation in architectural style and atmosphere. The "historic" buildings appear in a state of partial decay, with large sections demolished to make way for imposing brutalist concrete structures, characterized by their raw and cold aesthetic. To simplify these contrasting environments, the images are split between the two of them.

/ Scenery 01 -
Urban Remnents

Fragmented traces of the past, the beauty of the ordinary life.

Raw Renders

[SCROLL]
[SLIDE] > > >

/ Scenery 02 -
Fragmented Vision

A harsh, reconstructed landscape of brutalist forms, embodying a stark and oppressive modernity.

brutalist solitude
[SCROLL]
[SLIDE] > > >
torn street
[SCROLL]
[SLIDE] > > >
courtyard
[SCROLL]
[SLIDE] > > >
Due to complexity and performance limitations, the majority of the project's content is not available on mobile.

Curious? Check out the full page on tablet or desktop devices!
//Lazy Load Videos