Introduction
Every year, millions of devotees gather to witness the grand Jagannath Rath Yatra, one of India’s most spiritual and visually breathtaking festivals. Watching thousands of people pull the massive chariots of Lord Jagannath, Balabhadra, and Subhadra is an unforgettable experience. Recently, AI video generation has made something incredible possible—you can place yourself into a realistic Rath Yatra video without filming anything in real life.
When I first tried generating this type of video, the results weren’t great. Sometimes the AI completely changed my face, sometimes it created a different person altogether, and in many attempts the character was hilariously pulling the chariot in the wrong direction. It looked fake and completely ruined the realism. After several rounds of testing, prompt editing, and experimenting with Omni AI, I finally created a prompt that keeps the face locked, preserves the original body motion, and generates a believable Jagannath Rath Yatra scene. In this article, I’ll show you exactly how I did it so you can recreate the same result.
What You’ll Need Before Starting
The process is surprisingly simple. You only need three things:
- A clear front-facing reference image of yourself.
- A reference video with good walking motion.
- Omni AI video generation model.
- The detailed prompt that instructs Omni exactly what should and shouldn’t change.
The biggest mistake most beginners make is uploading only a selfie and writing something like “Make me pull Jagannath Rath.” AI fills in all the missing information by itself, which usually leads to identity changes, incorrect movement, unrealistic body animation, or strange camera angles. The secret isn’t using a better AI model—it’s giving the model extremely specific instructions.
Step 1: Choose the Right Reference Video
This step matters more than most people realize.
Instead of searching for an actual Rath Yatra clip, I used a simple reference video of a person walking naturally. The walking speed, posture, and camera movement were already realistic, so Omni only needed to replace the environment and integrate the Jagannath Rath into the scene.
A stable walking video works much better than a dramatic action clip because the AI can preserve natural body movement. If the original video already contains smooth walking, your final Rath Yatra video will also look smooth.
Step 2: Upload Your Reference Image
Next, upload your portrait image into Omni.
I always use a high-quality image where my face is clearly visible. The better the reference image, the easier it is for Omni to preserve facial identity.
In my prompt, I specifically instructed Omni to use my uploaded photo as the only facial identity reference. I also told it not to beautify, reshape, blend faces, or generate a different person. This is one of the most important parts of the prompt because AI models often try to “improve” faces unless they’re explicitly told not to.
Step 3: Lock the Identity
This is the section that made the biggest difference in my results.
Instead of writing a simple sentence like “keep my face the same,” I used a strict identity lock. The prompt clearly tells Omni to preserve my facial structure, beard, hairstyle, eyebrows, nose, lips, jawline, skin tone, age, and natural proportions exactly as they appear in my uploaded image.
I also instructed it to avoid face drift, morphing, flickering, or changing expressions throughout the video. Without these instructions, many AI-generated videos slowly transform the person’s face as the video progresses.
Step 4: Lock the Original Motion
This was another issue I kept facing.
Whenever I simply asked Omni to create a Rath Yatra scene, it generated completely new walking animations. Sometimes the character walked sideways. Sometimes the rope floated in the air. In a few attempts, the character was actually pulling the chariot from the wrong direction.
The solution was surprisingly simple.
Instead of asking Omni to create new movement, I instructed it to preserve the original reference video’s body animation exactly. The walking speed, posture, arm movement, footsteps, head movement, and camera motion all remain unchanged.
This allows Omni to focus only on replacing the person’s identity and transforming the surroundings into the Jagannath Rath Yatra instead of inventing entirely new body mechanics.
Step 5: Give Clear Rope Instructions
This is probably the most overlooked part of the prompt.
AI doesn’t automatically understand where the rope should be connected. If you don’t specify the direction, it often creates impossible physics.
To fix this, I clearly described the position of the Rath and the rope.
The Rath stays behind the character.
The ceremonial rope starts from the front of the Rath and extends forward into both of my hands.
The rope remains under natural tension throughout the walk, making it obvious that I am pulling the Rath while walking forward.
These few sentences completely changed the quality of the output.
Step 6: Build the Festival Atmosphere
Once the motion and identity were fixed, I focused on making the environment feel authentic.
Instead of simply mentioning “Jagannath Rath Yatra,” I described the complete scene.
The prompt includes thousands of devotees, flower petals falling from the sky, traditional Odia decorations, saffron flags, temple bells, conch shells, devotional chants, golden sunlight, cinematic lighting, and realistic dust particles.
These details help Omni understand the mood rather than just the location.
The result feels much closer to real documentary footage than a generic AI-generated festival scene.
Step 7: Add Natural Dialogue
One feature I really enjoy in Omni is its ability to generate synchronized speech.
Rather than using dramatic or overly emotional dialogue, I wrote a simple devotional sentence in Hindi that feels natural for someone participating in the Rath Yatra.
Keeping the dialogue short also improves lip synchronization. Long speeches often create unnatural mouth movements, while simple devotional lines usually produce much cleaner results.
Tips for Better Results
After generating many versions of this video, I noticed a few patterns that consistently improve quality.
Always use a sharp reference image with good lighting.
Avoid blurry or low-resolution selfies.
Choose a reference video with stable camera movement.
Keep your prompt detailed but organized into sections like Identity Lock, Motion Lock, Scene, Camera, Audio, and Dialogue.
If the AI changes your face, strengthen the identity section instead of rewriting the entire prompt.
If the rope behaves incorrectly, make the rope direction more specific instead of changing the scene description.
Small prompt adjustments often make a much bigger difference than switching to another AI model.
Common Mistakes to Avoid
Many people think a short prompt is enough, but that’s usually why the results look unrealistic.
Avoid using vague instructions like “Make me in Rath Yatra” or “Pull Jagannath Rath.” These leave too much room for the AI to guess.
Another common mistake is asking the model to create a completely new video instead of preserving the original reference motion. New animations are much harder to control and often lead to unrealistic body movement.
Finally, don’t overload the prompt with dozens of cinematic buzzwords. Focus on clear instructions, realistic movement, and identity preservation.
Final Thoughts
Creating realistic AI videos isn’t only about using the latest model—it’s about learning how to communicate with it. Omni is already capable of producing impressive cinematic videos, but the quality depends heavily on how well your prompt explains the task.
By combining a clear reference image, a stable reference video, strict identity locking, motion preservation, and detailed scene instructions, I was able to create a Jagannath Rath Yatra video that genuinely looks like I was present at the festival. If you’ve been struggling with changing faces, incorrect rope direction, or unnatural body movement, try following the workflow shared in this guide. With a few careful adjustments, you’ll be surprised at how realistic your AI-generated devotional videos can become.