Christopher Walkman, another experiment with contact sheet prompt chaining. New settings, new subject and scene. All prompts, workflow files, links, everything below.

My objective here was to prove that the setting and subjects can be way different to what I and others have posted. The key thing that makes this workflow pop is a central subject (or subjects) and sweeping dynamic camera motion.

This was again using Node Banana for the workflow, the whole workflow including all reference images, thought process etc is in the examples directory. You can just drag and drop it in.

If you're new to my page, I use contact sheet prompts a lot. This is a single prompt that generates 6-9 images in one hit. This way you can minimise conceptual drift across generations, and then extract/enhance each image. I wrote about it here:

Here's the contact sheet prompt for Chris: Analyze the input image and silently inventory all design-critical details: the model's features (older male, grey hair, serious expression), the styling (tan dickies work jacket, blue cargo jeans, vintage blue Sony Walkman, over ear vintage headphones), the dog, and the environment (trees, dappled sunlight). All styling elements, the model, environment details, and color grade must remain 100% unchanged across all frames. Do not add or remove anything. Do not reinterpret materials or colors. Do not output any reasoning. Perfectly replicate the exact structural features of the clothing, specifically the jacket texture, the plastic sheen of the Walkman, the classic over ear walkman headphones, the park setting, and the details and appearance of the dog. Your visible output must be: One 2×3 contact sheet image (6 frames only). The six frames must be spatially dynamic, non-linear, and visually distinct. Required 6-Frame Shot List 1. The Hero Profile (35mm, Eye-Level, Side Profile) Camera positioned at eye level, capturing the model in a strong profile view looking off-camera. The focus is on the silhouette of the bomber jacket and the over ear walkman headphone cable draping down. The background shows the depth of the park trees. The lighting creates high-contrast shadows on the face, emphasizing the serious expression. 2. High Angle (32mm, downward tilt) Camera positioned overhead but off-center, capturing the subject and dog from a diagonal downward angle. This frame should create strong shape abstraction and reveal wardrobe details from above. 3. Low-Angle "Power" Stance (24mm, Worm's Eye View, Upward Tilt) Camera positioned very low to the ground, angled obliquely upward. This elongates the model's legs and makes the jacket look imposing and architectural against the tree canopy and sky. The classic walkman over ear headphones are visible around the model's neck. The Walkman clipped to the belt is a focal point of the composition. The dog is visible in the shot, looking directly into the camera lens with a blank expression. 4. Wide Cinematic Environmental (50mm, Wide Shot, Center Framed) Camera positioned back, capturing the model and dog centered in the frame. This shot emphasizes scale and the isolation of the subject and his dog. The model stands hands-by-sides, looking directly at the lens. The dog looks directly at the lens of the camera with a blank expression. The sunlight cuts through the trees, creating a dappled light effect across the whole scene. 5. The "Listening" Mid-Shot (85mm, Chest Up, Shallow DOF) Camera positioned at chest height, slightly off-center. The model has one hand raised, adjusting the headphones. The focus is sharp on the face and the collar of the jacket. The background is compressed and blurred (bokeh), turning the leaves into washes of saturated color, but remaining recognizable as the park. 6. Waist-Level textural Detail (50mm, Hip Level, Medium- Close) Camera positioned at hip level. A focused study of the jacket hem, the denim texture, and the Walkman. The model's hand is resting near the device. The background includes the park setting behind the subject to maintain context. Continuity & Technical Requirements Maintain perfect subject fidelity in every frame: exact outfit details, Walkman model, and environment. Visual Style: Film Stock: Fuji Velvia. Color: Highly saturated, rich greens and reds, deep blacks. Texture: Pronounced, realistic film grain. Lighting: Hard, natural sunlight with strong contrast (sunny 16 rule). Lens: Varied focal lengths as specified. Output Format A) 2×3 Contact Sheet Image (Mandatory) Do not include any notes or text between the images. Do not show a significant gap or border between images.

Here is the dog mode one: Analyze the input image and silently inventory all design-critical details: the specific brown and white Brittany-type dog, its tan collar and yellow tag, and the environment (mature trees, sun-dappled green grass, fallen autumn leaves, hard natural sunlight). The subject is exclusively this dog. Crucial Constraints & Instructive Elements: All aspects of the dog (appearance, collar, tag), the environment details, lighting conditions, and the specified color grade must remain 100% unchanged across all six frames. Do not add or remove anything. No human models, no props, no extra clothing, and specifically no headphones or Walkman devices should be present. Do not reinterpret materials or colors. The fur texture, grass, and leaves must be rendered exactly as seen in the input image. Perfectly replicate the exact structural features and appearance of the dog and the park setting in every shot. Maintain perfect subject fidelity in every frame. Visual Style: Film Stock: Fuji Velvia aesthetic. Color: Highly saturated, rich greens and reds, deep blacks. Texture: Pronounced, realistic film grain throughout the entire contact sheet. Lighting: Hard, natural sunlight with strong contrast (creating distinct shadows), matching the input image. Output Format: One single 3x2 photographic contact sheet image containing 6 distinct frames. The six frames must be spatially dynamic, non-linear, and visually distinct perspectives of the same dog in the same moment. Do not include any notes, text, film strip numbers, or markings between or around the images. Do not show a significant gap or border between images; they should sit tightly together as a single sheet. Required 6-Frame Shot List (Grid Layout): Top Row (Left to Right): 1. The Hero Profile (35mm, Eye-Level): The camera is positioned at the dog's eye level. The dog stands stoically in a strong side profile, looking off-camera right. The hard sunlight highlights the contours of its face and fur. 2. High Angle (32mm, Downward Tilt): The camera looks down from above. The dog is looking up towards the lens. This angle emphasizes the shape of its body against the textured ground of grass and fallen leaves. 3. Low-Angle Stance (24mm, Worm's Eye View): The camera is very low to the ground, angled sharply upward. The dog looks imposing, towering over the lens against the backdrop of the large trees and sky. 4. Wide Environmental (50mm, Center Framed): A wide shot capturing the entire dog centered in the frame, emphasizing its scale within the large park environment. The dog stands alert, looking directly at the camera. 5. Portrait (85mm, Close Up, Shallow DOF): A tight close-up on the dog's face. The focus is sharp on the eyes and nose. The background of sun-dappled leaves is heavily blurred into bokeh. 6. Textural Detail (50mm, Hip Level): A focused study of the dog's shoulder and flank texture, showing the coarse fur and the details of the tan collar and tag. Sunlight rakes across the fur texture.

For the i2v step, just use Kling 2.6. I know it's only in the webUI, but it's just way better. The key thing to smooth motion that I don't see others doing is to choose key frames that have visible overlap, and to prompt for smooth motion. Here's one of my i2v prompts: "The camera slowly and smoothly pushes in while lifting on a boom to highlight the models face. The text 'CHRISTOPHER' appears in the background in a 3d folding animation. The model slowly lifts his arm to touch the headphones on his head. The model is extremely slow and delibeerate in his movements. The dog is completely still. The park is completely still on a windless day." Here is a referral code, if you want to hook me up with some free tokens. I am not affiliated with Kling:

The key thing here is strict control over the motion of the subject. Any fast motions come across really stuttery when the ease curve is applied. You can see a bit of this in the dog sequence. If I wasn't lazy I'd have smoothed that out in the i2v prompts.

Then for stitching, ease curves, and background music ; use (another one of my projects I made just for this)

What I love about this is that the lens perspective also changes. Ie.: when wide angle, field of view changes; when tele, fov changes again 👍

You can get some heavy dolly zoom effects going if you prompt for it. I'm into that.

this is dope!

Thanks!

This goes harder than it has any right to

💪

I like what you did here with "Walkman" 🤣

ngl, I expected to see this comment sooner 😂 I appreciate your appreciation

All good things come to thos who wait :P

Really good work

awesome stuff

Look at the leaves. AI stands for Artificial Inconsistency

👶

nice ! and he's my favorite actor

One of the all time greats for sure

Bit hard for beginners

Get an LLM to walk you through it, would be a good opportunity to learn some new things

Willie, what would you say is the average cost for these videos you've been making? $5-10 or higher?

Hey Greg, the cost for the image gen workflow is just under $5, but I probably spent $20-30 while building it.

Just for the images, not including the videos?

Yep, just the images. For the videos I used Kling over the web UI on a paid plan. Not sure what that cost works out to.

But images are free with Nano Banana using the Gemini API key, no? Mine were free. I used Higgsfield for the videos, seems the cheapest option right now... 200 Kling 2.5 videos per month (6 credits per video) compared to 120 on the Kling web UI (25 credits per video)

But I love kling 2.6 first+last frame and hate money. Also you'll hit a wall with the gemini API key,

I sent you a DM

I've tested new settings; results varied.

How so?

Amazing

🤟🤟

this is amazing, great job

Thanks behnam 🙏

This is awesome 😎

Big up morris

I understand video compiling, I just don’t know how to get set up with that GitHub stuff.. I’m basically cooking on my kid’s iMac 🖥️

Cool, the video really in a 1 time or separate?

Yes. The hundredth plug and play variation of the same idea. Over and over and over. Grow some balls and do one with Jeffery Epstein.

@mk_beta_kitten I thought about it, but it's a fine line between parody and pedophile hype video

Hi @ReflctWillie do you have an youtube where you explain this process, I'm new into this and I'm completely lost.

I don't sorry... You could check the recent post from @VibeMarketer_, it follows one of my older workflows to the letter, but with Higgsfield instead of Node Banana. But I reckon you should have a go at getting Node Banana working, you will pick up some new skills along the way! Instead of me writing out a step by step. Go to chatGPT and paste this prompt into the chat: "I want to install and run this project: Assume I have no experience with installing projects from GitHub, using basic terminal commands, or any of the required software installed, e.g. Node/NPM. Keep the guide straightforward and succinct, include links to any pages where I need to install things. Include steps for getting a Gemini API key from Google AI Studio. Do a web search to ensure that you have up to date links and information before writing the guide." Feel free to edit that to your skill level

@VibeMarketer_ thanks alot

am I just that crazy, they could not be that happy about it? :S ot this is all the way "etichs" ? it will blow up in 1 week!

Love the enthusiasm

Maybe I was not clear. I might do this, but i feel it should be a famous person to be "viral" but I would not use anyone for that because of ethical reasons. I will most likely try this on someone "artificial". Have a great one.

If you're using it for commercial purposes then go with an artificial person, or someone you have talent rights for. Otherwise you'll get in trouble.

Saved, will go through this in free time. Great work 🔥

Hmu if any questions once you get into it

Hey @ReflctWillie this was really good. Only question I have is the quality for the contact sheet is decent enough to use? If cropped? In my line of work I need to generate multiple angles of the model to animate and make ads for it. So I can definitely try this method

There is a grid splitting node in node banana for this purpose Basically generate the sheet, then upscale and reframe in another call

That's helpful. Will try this one. Thanks.

It needs a little more, doesn't it?

Took me a sec

👀

wouldn’t this be achievable with getting different angles either from higgsfield shots, nano prompts, etc and then using start frame end frame in kling?

Yeah for sure you can replicate all the same steps in other programs. I'm building my own thing because I want to.

whoa

This is amazing. Isn’t the Gemini API expensive ? What if we create images saperately ?

Love the result! Appreciate you going extra mile and explaining things in detail 😌

I appreciate it, I enjoy writing about my work/process.

Can you don for me

@Badoo__X I assume you are referring to donburi here? If so, hell yeah. I'll don any time any place.

This is awesome!

Cheers Dan!

Awesome! 😉👍

Honestly, I wouldn’t want to be an actor these days - they can do whatever they want with your face using AI 😅

I'll do you next