This video really tested my patience in both getting runway to create what I needed. The videos it spat out always had the issue of moving the camera by zooming in a lot when I wanted it to be mostly stationary despite checking the stationary camera box (cause if I check that, the woman does not move at all).

The lip syncing part was pretty start forward and fun, I used character.ai in order to get a female voice that matches to the AI woman, I actually like the fact that she sounds a bit robotic as it automatically creates the feeling of things being fake and not creating a connection for the man.

Lastly, getting the video down to 1 minute (1:08) was a genuine challenge, putting in all of the video I captured I had 8 minutes, my first edit down was 3 minutes, then 2, then 1:30. I finally got it down to 1:08 and knew that if I cut anymore, it would impact the story of the video.

Here are some screenshots of the work that went into creating this video and the AI prompting for getting the images, videos, and even the AI voice using character ai:

Midjourney’s Mid Journey

Spent 3 days trying to find the exact look I want. It took quite some time to get it close to what I envisioned. Though I am a bit of a perfectionist, so I could’ve settled for earlier iterations. I ended up sort of doing that anyway.

Style References

Used a couple of these as a style reference. However, I noticed text prompting was easier than styling – Especially if I attached more than 1 style, I’d get some weird results and the aliens I had established would change entirely.

Top Alien Contenders

I really like the detail in the first picture, but had trouble translating that to a background setting that looked like it would be a YouTube video and not 60 Minutes. Also had trouble getting them to smile. So, I used the first picture as my main reference for the other aliens.

Current Winner

I picked this one as my main characters, since the aliens and the background seem like it would be in a YouTube video. For a while, I’d either get a really good background but the aliens look the exact same or had no special detail or character to them. I’d also get the opposite, where the aliens look very personable and different from each other, but the background was plain. Instead of a bedroom, it looks like they’re in a low budget set, which works for these characters.

I used Photoshop’s generative AI to expand the image a bit and change their hands. In the above 10 photos, this original one, straight from Midjourney, had some stretched, morphed hands.

I might change the image again, as these two aliens are not looking in the same direction and I’d need a bit more room on the right side to follow the rule of thirds, as the PIP reactionary video would probably sit on the right bottom corner.

Scott McCloud’s Understanding Comics provides a fascinating framework for understanding how time is constructed, not just in comics, but also in cinema. McCloud argues that while comics present a “time map” where past, present, and future are simultaneously visible, cinema primarily operates in a sequential “now.” However, skilled filmmakers can manipulate this “now” to create subjective experiences of time, mimicking some of the effects McCloud attributes to comics. A prime example of this can be found in “Fishes,” the intensely stressful Christmas episode of The Bear (season 2, episode 6).

McCloud emphasizes that panel size in comics influences the perceived duration of time. A wide panel suggests a longer timeframe, similar to a long cinema take. “Fishes” utilizes this principle during the climactic confrontation between Mikey and Uncle Lee at the dinner table. The extended, often unbroken takes of their escalating argument stand in stark contrast to the episode’s generally frenetic, quick-cut style, which mirrors the chaotic preparation of the seven fishes dinner earlier in the episode. These long takes, focusing on the raw, volatile emotions, stretch time, forcing the viewer to inhabit the agonizing discomfort of the moment. We’re trapped at that dining table with them and Carmy, experiencing the emotional weight of every insult and accusation thrown.

Furthermore, the concept of the “borderless panel” in McCloud’s work is that it evokes timelessness. This episode uses a different method to play around with our sense of time, it makes use of a “ticking clock”. The entire episode is built around a relentlessly advancing “clock”: Donna’s cooking. Her increasingly erratic behavior and the impending deadline of dinner serve as a constant, driving force. This is almost the opposite of a borderless panel, an underlying countdown that gives form to the entire episode. Creating a “time map,” similarly to flipping pages in a comic and feeling like you know the weight of the impending future and accumulated past. Every slammed pot, every frantic stir, every shouted instruction tightens the tension, creating a subjective experience of time accelerating towards an inevitable explosion. Donna is the timer.

Ultimately, “Fishes” demonstrates that while cinema lacks the spatial “time map” of comics, it possesses its own powerful tools for temporal manipulation. By employing long takes, a relentless “ticking clock,” and stylistic shifts, the episode creates a visceral, subjective experience of time, mirroring and expanding upon many of the techniques McCloud details in his exploration of comics. The episode does a great job in using cinematic language to not just show time, but to make the viewer feel it.

Time manipulation in digital cinema is critical in making the viewer feel the true action and pacing of a scene.

In ‘Shaun of the Dead’, director Edgar Wright creatively manipulates time to make viewers entertained while seeing the same scene several times. While Shaun is coming of with ‘The Plan’ for how they will survive the apocalypse, he amends it each time, while keeping the core plan the same. However, rather than completely skipping over showing the plan multiple times, Wright utilizes the same shots each time, sped up or slightly changed. This adds a great comedic effect to the scene and allows viewers to glimpse the quick montage multiple times, hopefully capturing different details in each pass.

This montage unfolds as a quick sequence of events, further highlighting that this is an imagined scenario rather than real life. This is aided by the vocal narration, which furthers the urgency felt and the similarities of each pass of the montage. Overall, it’s an extremely effective and fitting way to present a plan of future actions that not only keeps the viewer entertained, but it also adds to their experience despite seeing the same footage multiple times in a row.

(Best comedy trilogy!)

The film begins with mostly still nature shots. This is slowly interspersed with shots of the woman as she reads a letter and has an emotional reaction. The juxtaposition of the emotional shots of the woman with the landscape shots draws the viewer deeper into her mind, as the mood of each scene carries from one to the next. The viewer can see that she is upset, and scenes with darker emotion such as the heavy rain push the feeling further. There is also a sense of death or loss, as the trees lose their leaves and stand barren. While the viewer can try to track a sense of movement or action through the piece, overall, the only focus is on the woman and her feelings compared to the nature around her.

‘Autumn Mists’ style of montage is very effective, as it conveys feelings without words in a way that facial expressions alone could not. Using the Kuleshov effect, the emotions and expressions that could be misunderstood are further defined through the non-human imagery. It evokes a feeling of sadness and heartbreak, but even further than that, it evokes a feeling of hopelessness and despondence. She almost seems to be in shock, although that could also be attributed to the style of acting at the time. However, the comparative shots of the woman to the nature around her are quite moving and will speak differently to each viewer.

This was the shot list for the second idea I had for the AI project and is no longer the current plan for the assignment. At this time, I didn’t have any reference images for style.

——————————————————————————————

Shot 1: The AI Awakens

• Description: We open in complete darkness. A burst of abstract digital light forms into a spherical shape. The AI’s form is fluid, not humanoid—think glowing spherical patterns, constantly shifting and expanding. It doesn’t have a defined body but seems to float and pulse gently. The space around it remains empty, suggesting its initial, undefined state.
• Style: Spherical, glowing digital formation. The color palette starts in grayscale, with hints of digital static or noise as it assembles.
• Duration: 5 seconds
• Tone: Quiet, introspective. The AI’s awakening is subtle and almost alien, suggesting it’s in a state of curiosity.

Shot 2: The AI Observes Sadness

• Description: The AI begins to witness sadness, but not in the form of figures—rather, through fragments of life. A scene of rain hitting a window, perhaps a slow-motion drop of water running down glass, or a wilting flower, or an empty swing swaying in the wind. These objects subtly convey sadness.
• Style: The imagery of rain, faded colors, and slow, deliberate motion. Everything feels detached and distant, as if the AI can only observe from a distance. The colors are muted—blues, grays, and darkened tones.
• Duration: 7 seconds
• Tone: Pensive, detached—the AI senses sadness but doesn’t feel it, only observing the decay or emptiness of the moment.

Shot 3: The AI Encounters Joy

• Description: The AI sees moments of joy, but still in the form of fragmented imagery. A balloon floating away in the sky, sunlight reflecting off a puddle, or birds soaring through the sky. The camera captures these fleeting, beautiful moments—small and spontaneous.
• Style: Bright, energetic bursts of color—yellows, oranges, and light pastels. The movement is quick, the visuals vibrant and alive. There’s an abstract quality to the way joy is depicted; it’s present but not fully grasped.
• Duration: 7-8 seconds
• Tone: Wonder, brightness—the AI is intrigued, fascinated by this energy, but remains distanced.

Shot 4: Contradictory Emotions

• Description: The AI witnesses the contradictions of human emotions: a shot of someone laughing while holding back tears, or a cracked mirror with fragmented reflections. A flower blooming in an empty, abandoned place. The AI’s perception of joy and sadness happening simultaneously, represented in seemingly incongruent fragments of life.
• Style: This shot is visually chaotic, with multiple split-second glimpses into different emotions—broken, fragmented, and chaotic. The visuals may have overlapping images, with the AI’s spherical form subtly pulsing in the background as it tries to make sense of these contradictions.
• Duration: 7 seconds
• Tone: Overwhelmed—the AI sees the contradiction and tries to understand, but it only results in confusion.

Shot 5: The AI Sees Connection

• Description: The AI sees a subtle, quiet moment of connection in the world—a hand reaching out, an embrace, or a shared smile between strangers. But rather than seeing the people themselves, it’s the subtle details—the movement of a hand, the soft curve of lips, the warm glow of a sunset in the background.
• Style: This shot is calmer, more harmonious. Soft light, gentle movements, and natural colors (peach, warm golden tones, soft pastels) suggest a peaceful connection. The AI floats through the scene, abstractly observing these moments.
• Duration: 7-8 seconds
• Tone: Reflective—the AI sees the beauty of connection but is left slightly confused by the depth of the emotion.

Shot 6: The AI’s Reflection (Ending)

• Description: The AI observes a peaceful moment, perhaps something serene—like the stillness of a sunrise or the quiet ripples on a calm pond. It floats in place, the world around it peaceful and still, but there’s a subtle, lingering distortion in the visuals—like a glitch or flicker that suggests the AI’s inability to fully grasp the experience of emotion.
• Style: A peaceful visual with slight digital distortion—a soft ripple effect on the water, subtle glitches in the light. The colors are calm, with pastel tones and soft gradients.
• Duration: 5 seconds
• Tone: Thoughtful, introspective—although the AI is observing beauty and calmness, there’s still a quiet sense of distance and incompletion.

My current idea, which I will plan out with a shot list and reference images after I complete my missing projects, goes like this. It’s a short comedy that follows two scientists working in a secure and secluded lab together. One is looking through a microscope and discovers something troubling and asks the other for their input. As they each go back to look at the thing it gets weirded causing them to consider the panic button and chaos ensues.

Here is a sort list I have of static image generation. I have run all of these through Adobe’s AI service in Photoshop just to get an idea for style and what the shot might look like. Some of these have varied styles, but I am going for more of the yellow photograph style, and moving away from the illustrated style. Here are my shot descriptions and some of the photos I generated. As I begin the process of working with the video service, I will alter the prompts to add movement in the shots.

Shot 1

give me a photo of a cowboy entering a saloon with swinging doors. This cowboy will have his face hidden under a cowboy hat, and the saloon will have many guests all turned towards the cowboy. This photo will be in the old west style, very yellow and in the style of an old photograph

Shot 2

give me a photo of people from 1882 playing cards at a table in a saloon from 1882. The players should be playing poker with a classic deck of playing cards, and on the table there should be money, pokerchips, and alcoholic drinks. This photo will be in the old west style, very yellow and in the style of an old photograph

Shot 3

Give me a photo of a bartender from 1882 making drinks in a bar with lots of guests. The guests are on the other side of the bar, and the bartender is giving them drinks. Focus on the bar tender. This photo will be in the old west style, very yellow and in the style of an old photograph

Shot 4 

Give me a photo of people sitting at a table in a bar from 1882 looking up from their drinks. Have them looking at the camera. This photo will be in the old west style, very yellow and in the style of an old photograph

Shot 5 

Give me a photo of a cowboy entering a bar in 1882 with swinging doors. This cowboy will have his face hidden under a cowboy hat, and the bar will have all the guests all turned towards the cowboy. Have the camera focus on his hand as it reaches for his pocket. will be reaching for something in his pocket This photo will be in the old west style, very yellow and in the style of an old photograph

Shot 6 

Give me a close up on a cowboy’s face as he smiles.  This photo will be in the old west style, very yellow and in the style of an old photograph

Shot 7 

Give me a cow standing in a bar from 1882. This cow should have a gun and be wearing a cowboy hat. This photo will be in the old west style, very yellow and in the style of an old photograph. Have him be half human with human limbs. Show full body

Shot 8 

Give me a cow standing in a bar from 1882. This cow should have a gun and be wearing a cowboy hat. This photo will be in the old west style, very yellow and in the style of an old photograph

Shot 9 

Give me a photo of a bartender from 1882 making drinks in a bar. The guests are on the other side of the bar, and the bartender is giving them drinks. Focus on the bartender. The bartender is pouring a glass of milk. This photo will be in the old west style, very yellow and in the style of an old photograph

Story Summary: I was thinking of taking inspiration from a YouTube creator, Neural Viz, who uses AI as a tool for creativity rather than an empty crutch. Their work uses AI to enhance their comedic meta-commentaries, and I want to aim for something similar – a spoof that uses AI to make a funny observation about online video culture itself. My idea is to create a spoof YouTube reaction video. Imagine a typical, slightly messy room, the kind you see in countless reaction videos online. In this room sit two aliens, Zorp and Gleep, who are set up to film a reaction video. The main screen behind them shows a small picture-in-picture (PIP) display. This PIP is where the “content” they are reacting to will play – and that content is intentionally terrible AI-generated video. Think abstract blobs of color that are supposed to be landscapes, nonsensical AI animations, or AI-generated human figures with constantly shifting, uncanny faces. The humor will come from the aliens’ over-the-top reactions to this objectively bad AI art. One alien, perhaps Zorp, could be completely enamored with the AI video, praising its “deep artistic expression” and “groundbreaking visuals,” while the other, Gleep, is utterly repulsed, making exaggerated gagging noises and complaining about the “atrocious” and “incomprehensible” imagery. The contrast between the truly awful AI video and the aliens’ strong, often opposing, reactions will be the core comedic element.

Visual Style: The visual style will deliberately mimic the low-budget, casual aesthetic of typical YouTube reaction videos. Think of channels like MoistCr1TiKaL, Cody Ko, or Danny Gonzalez, but also even more average, everyday YouTubers. The “set” would be a normal-looking room, maybe a living room or bedroom, slightly cluttered but not intentionally styled. Lighting would be practical and unflattering – overhead lights, desk lamps, nothing cinematic. Camera work would be simple, maybe slightly grainy, with basic framing, nothing too composed. Thumbnails would also follow the reaction video style – featuring the aliens with exaggerated facial expressions of shock, disgust, or amazement, maybe with some low-resolution text overlay. The key is to create a visual style that is instantly recognizable as a YouTube reaction video but with the added comedic element of alien characters and their over-the-top reactions to the intentionally jarring and low-quality AI-generated video playing in the PIP.

I want to create a film based on a saloon scene from the old west. I want to have a cowboy enter a saloon. The whole bar will stare as he enters, disrupting a card game. All of the patrons l will look terrified as he enters. There will be a shot of the bartender making drinks as everyone is focused on this man. He will then smile and reach for what looks like a gun, He will then begin to turn into a cow. He will stand there as a cow dressed in a cowboy outfit with a gun, a hat, and maybe dashing vest. The bartender will then make him a glass of milk. I want the style to be that of an old western, based off of an old photograph from the late 1800’s. I am fine if it is in black and white, but would like the color grade to have some yellow. I described the set as from 1882 to give the AI the style and time period I wanted. This is set in a bar with drinks and poker, and folks dressed as cowboys or normal garb from 1882. I am still trying to find interesting ways to expand the story while keeping the concept simple. I Want the expressions on the patrons of the bar to be scared and tense. I want the style to be like some of the scenes in Butch Cassidy and the Sundance Kid, with high contrast and warm yellows.

This is the second part of my ChatGPT conversations and composition of my AI Video project. For more early conception ideas refer to “Rylan Eisenhauer Blog Post #4 (Chat GPT Video Chat)”. This post is about the specifics in narrowing down the shots I want to create for the project using ChatGPT to edit and narrow down my ideas. Below is a screenshot of my finalized list of shots and expressed emotions that I am aiming for, and below that is the entire conversation I had with ChatGPT to edit and narrow down my ideas:

Since the video work I am creating consists of sad and slow shots, I should only have about 7-8 shots as I think the drawing out of scenes and panning shots can help extenuate the feeling of things being slow and gloomy. Too many fast and whippy shots can ruin the effect I am going for as well as give the audience “too much” information when things should feel a bit ambiguous so that many can relate their experience to my video.

I was recommended to try and practice using RunwayML a little bit in order to experiment with AI creating video. I knew exactly what I wanted to test as I have had a still graphic of a lens flare that I wanted to give some motion. I gave RunwayML the image with the prompt:

“the yellow/orange flare in the middle does not move from the middle, but the rays that it shines slowly shine outwards without covering the red background. The video should end looking how it started as if it loops.”

Here is the still image as well as the video it gave to me as a result of the prompt, it was almost exactly what I was looking for besides the loop part:

This is exactly as I was looking for and I am glad it seems I understand that being specific with prompting is important. I will ensure to keep this strongly in mind when making my AI video and ensure that there is a strong sense of consistency in the videos as well as characters (such as the female AI character) looks consistent and natural.

-Rylan Eisenhauer

Kirsanoff’s Brumes d’Automne is a profound vignette of a woman’s sorrowful, pensive state. It begins with shots of bodies of water, towering trees, and overcast skies, setting a melancholy tone that persists throughout the film. The rain picks up as the young woman sits by her burning fireplace and opens a letter. Disappointed by its contents, she places it in the fire, letting it burn. As her eyes drift from the blaze, she slowly looks into the camera, and a methodical blurring effect signifies that the scene is part of her memory.

She stands in a doorway, watching as a man—presumably her former partner—walks away. The scene cuts between shots of her staring longingly, the gate outside the house, and the man’s legs as he leaves. This sequence effectively captures the source of her sorrow. Following this, the montage style continues, showing the letter burning and charring, the rain intensifying outside, and her eyes glossing over as she holds back tears. The interplay of these images evokes a deep sense of melancholy that swells as the memory overtakes her.

The woman leaves her home, wearing a shawl and heels as she walks through mud and puddles. Around 7:45, the editing becomes especially impressive for its time. As she gazes over a large body of water, the film manipulates perspective, twisting the view to encapsulate her mystified state. The landscape continues to warp, indicating that she “can’t see straight.” The film is remarkably effective in allowing the viewer to understand and connect with the woman’s frame of mind. The combination of montage editing and natural scenery evokes far more emotion than if the film had relied solely on continuous editing.