How to Create AI-Generated Video in a Day (Adobe Express + ChatGPT Workflow)

Originally created by @dailywebdesign.

What if you could turn a single photo into a cinematic promo video, no photoshoot, no film crew, no training footage, in one day? That's exactly what @dailywebdesign did for his friend Julian, a kickboxer preparing for a fight. Using ChatGPT for scripting, Adobe Firefly for visual generation, and Adobe Express for content assembly, the entire workflow went from idea to finished video to educational carousel in a single session.

Here's the step-by-step breakdown.

---

Step 1: Start with a Concept and a Single Photo

The project started with just one photo of Julian and his fight announcement, "Kelch vs Feldman, 170 lbs Kickboxing Bout." No professional photoshoot, no training footage, nothing. Just a face and an idea: tell the story of Julian preparing for fight night.

The key insight here is that you don't need expensive equipment to create compelling video content. You need a clear concept and a reference image. Everything else can be generated.

---

Step 2: Use ChatGPT to Build a Scene-by-Scene Storyboard

The next step was describing the story to ChatGPT: Julian hitting the bag, jumping rope, working through push-ups, sparring, preparing himself for fight night. ChatGPT helped transform that raw idea into a scene-by-scene visual storyboard with detailed shot descriptions.

The storyboard included specific shot types, camera movements, and atmospheric details:

• Establishing shot: A wide, medium shot with hand wraps, gloves, and a gym environment. Serious, minimalist, dark aesthetic with soft lighting.
• Opening (extreme close-up): The fighter's tightly wrapped fist near his face. Sweat drips onto the bag. A slow, deliberate camera movement communicating exhaustion and determination.
• Round two (close-up): An intense close-up of the face. Skin covered in sweat, breathing heavily, eyes focused. The camera feels close and personal.
• Round three (medium): A tight side-angle as he begins striking the heavy bag. Zoom in to gloves upon impact, then pull back to reveal the full scene.
• Impact shot: Powerful, dynamic punches alternating between close, medium, and wide angles. The camera pushes in during each punch and pulls out to make the action feel cinematic.

This level of detail in the storyboard is what makes the difference between generic AI video and content that feels intentional and directed.

---

Step 3: Generate Visuals in Adobe Firefly

With the storyboard and Julian's reference photo in hand, the next step was Adobe Firefly. Using the scene descriptions and the reference image, the workflow generated:

1. Character sheets, multiple angles and expressions of Julian to maintain consistency across all generated frames.
2. Training session visuals, each moment from the storyboard rendered as a still image, keeping Julian at the center of the story.

Character accuracy was the priority here. By feeding Firefly both the shot descriptions and the reference image, the generated frames stayed visually consistent . Julian looked like Julian in every shot, not like a random AI face.

---

Step 4: Turn Frames into Video Sequences

Static images aren't a video yet. The next step used Adobe Firefly's frame-to-frame feature to turn the generated images into short video sequences. Each still became a dynamic clip, the camera pushing in, pulling back, panning across the scene.

The shot sequences from the ChatGPT storyboard served as the prompts for each video clip. Once all the individual clips were generated, they were assembled and edited together into one continuous video.

This is where the specificity of the storyboard pays off. Generic prompts produce generic video. Detailed shot descriptions, camera angles, movement direction, lighting, atmosphere, produce clips that feel like they were shot by a cinematographer, not generated by an AI model.

---

Step 5: Turn the Process into Educational Content

Here's where the workflow gets meta. Instead of just sharing the finished video, the entire process was brought back into ChatGPT and connected to Adobe Express through MCP (Model Context Protocol). The conversation that produced the video became an outline. The outline became slides. The slides became the educational carousel itself.

This is the part that makes this workflow special, one idea became a story, a storyboard, a collection of visuals, a finished video, and an educational carousel, all from the same conversation, without starting from scratch.

---

The Full Workflow in One Day

Here's the complete workflow:

1. ChatGPT . Script the idea, build the storyboard, generate shot descriptions
2. Adobe Firefly . Generate character sheets and training visuals from the reference photo
3. Adobe Firefly (frame-to-frame) . Turn still images into dynamic video clips
4. Adobe Express (via MCP) . Turn the entire process into educational content

The philosophy: Create it. Break it down. Teach it. Repeat.

What idea has been sitting in your notes that you could finally bring to life?

---

Build Your Own AI Creative Workflow

If you want to build AI-powered creative tools and workflows without writing code, Base44 lets you generate full-stack web applications from a text prompt, pages, components, database, and deployment included.

And when you're ready to present your work, NewForm Studio lets you generate professional mockups in seconds. Upload your design, pick a device frame, and download.

---

Follow @dailywebdesign for more AI design workflows and creative content.