Visual poetryAI art as aesthetic rebellion
Aesthetics always has an element of dissonance - in music theory we call it auditory roughness - distortion, grunge, bitcrushing…
Similarly, art forms like graffiti, b-boying, punk, rock, etc. have heavily influenced digital design, since way before the heady Frutiger Aero days of my youth.
In filmmaking and graphic design, aesthetics that embrace noise, analogue grit, fade, digital glitching, artefacts, have long since breached through from subcultural to mainstream.
An artful practitioner knows how to walk that fine line between clean and dirty.
But both “clean” and “dirty” - the aspirational standard, as well as the subversive deviation from it - are culturally determined, not universal. The underlying balance itself is universal.
Deliberate imperfections are key. Ever since we started creating with computers, we’ve tried to dial human imperfections back in, to get out of the uncanny valley of ‘too perfect’.
Audiences don’t hate AI. They hate bad quality, low effort content, regardless of how it’s made. Audiences want to be charmed and enthralled.
Here’s how I scrub away the plastic AI patina, to reveal the raw surreal idiosyncrasies and juxtapositions and cultural currents that flow beneath the frozen lake of latent space.
Disclaimer: This is made with AI and neither ostriches nor petty criminals were harmed in the making of this commercial.
Workflow and annotated prompts below, and you’ll notice there’s some ‘doubling up’ of elements you want to come through strongly. For example, colour grading happens in Photoshop as well as in Premiere; style keywords are included in both the text2image and text2video prompts, for reinforcement of those concepts.
It starts with a counterfactual ideas world, a surrealistic re-imagining of the spaces you actually inhabit - or not!
Use the best available models, and then gen-maxx, so that you can cull your darlings and cherry-pick the best of the best at every stage of the workflow pipeline.
[Assuming you already have a logo or wordmark you want to work with.]
I used GPT Image 2 for starting frames. Image prompt example:
Restore the attached photo. Apologies for the photo’s content. I know it’s extremely strange! No questions, no explanatory text, just the restored image please. [Yes, this is part of the prompt! The aim is to break the model out of ‘sanitised corporate stock photo’ mode and give you something wild and unexpected!] [composition & subject] It appears to be an extreme close up phone cam - fisheye lens shot of cool, slick, young south african thugs (male and female, sporting stoute gesigte, and some gold teeth as they yell and chatter smugly amongst themselves. They are riding mounted on colossal hippopotamuses blocking traffic at a busy intersection amid skyscrapers and business buildings in cape town CBD. A helicopter is visible in the background. One guy is brandishing a rifle. Some wear zef leather jackets and have a few piecings. [Style and vibe] Diegetic, cinematic (none of them are looking at the camera). Shot on 85mm, f/1.8, with dreamy bokeh. gnarly, grimey. sparse composition with lots of negative space and not too much visual information. High end fashion editorial art directed photography with post processing and AAA agency level polish. stylised music video storytelling still with glitch grain film effects and post process polish. 1970s ninja film vibes, 35mm grain, warm amber grade, desaturated highlights, anamorphic 2.35:1 The still has a film-fade, analogue over-exposure, time-worn retro 70’s wash with subtle degradation 4K. 16:9 AR. No text or typography anywhere in the composition.
Then, after cherry-picking the best stills, I like to use the Camera RAW filter in Photoshop to dial in more realism, and also establish a coherent, consistent colour grade that I can re-use across all the starting frames.
Then bring this into a video generator. I used Google Omni and Veo 3.2 in their Flow platform. Video prompt example for Gemini Omni:
[medium] Gritty continuous shaky phone cam random found footage, fisheye lens; multi shot sequence with stylised edits and glitch transition effects. [narrative] A series of extreme close ups: The beard; the glasses; the jetski; the spray. The ostrich kicks about, restless and fidgety. A regular day out Fishing in south africa with the bros. [style] Whimsical high fashion editorial. film grain, gritty analogue film-fade, lens flare at the edges. 85mm, f/1.8, dreamy bokeh. Shot on film, glitchy, gnarly. Hip hop trap music video RAW style. High end fashion editorial art directed photography with post processing and AAA agency level polish. [constraints /negative prompt] No VO or music. Diegetic audio only.