Exploring OpenAI’s Sora: The Good, The Bad, and A Bison

AI-Generated Videos Included

Hello Video Stars,

Does AI kill the movie star? I don’t think so—AI’s role in the entertainment industry is much more complicated and nuanced than I have time to unpack today. But Sora is here, and while it’s impressive, there is still room for improvement.

Today, I had the chance to play with OpenAI’s Sora—something I, as a filmmaker, have been eagerly anticipating for a while.

I’m currently juggling a lot, so I decided to use my work breaks (thanks to the Pomodoro method) to dive into Sora. That said, this post comes with a warning: I haven’t fully immersed myself yet, so these are my initial thoughts.

I ran out of tokens before I could dive deeper, but I’ll do a follow-up once I have access again. For now, here’s what I discovered in my brief interaction with it.

First Impressions: Sora Shows Promise, But It’s Still Early

My initial take is that Sora definitely shows promise. However, it’s still in the early days, and there's a lot of room for growth.

Video Quality

Out of the 20 prompts I rendered today, only five were great. That’s a 25% success rate, which is decent but leaves room for improvement. The remaining 75% of outputs were far off the mark—characters looked strange, some had a green screen effect, and many people were walking awkwardly—lots of weird walking.

But here's the thing: when Sora got it right, it really impressed me.

The details, lighting, and overall quality of the five good videos were surprising. Those small touches really elevated the visuals.

Storyboard Function

Another area where I wasn’t impressed was the storyboard function. The lack of control left me feeling a bit disappointed. It could be that I haven’t spent enough time with it yet, or perhaps it’s still in development. Either way, it didn’t offer the flexibility I need, especially as someone who likes to have control over the finer details of a shot or sequence.

Preset Options

I’m in the middle of doing a deep dive into the presets. I’m excited to explore them further, but I didn’t get far enough today to feel confident with them. Stay tuned for more on that later.

The Results

My university is located in Nashville, and we are the Bisons (yes, with an s). I thought I’d have a little fun with some "A Bison on Broadway" video. Let’s take a look at the good and the bad. By the way, Sora generated the following titles for each video.

Let’s start with the good…

Bison Stroll on Broadway

This is one of my favorites. The detail in the fur, the vibrant colors of the sunset, and the way the light bounces off the street are impressive. I love how the cars slow down as the Bison casually roams by. The neon reflections on the car to the left of the screen are a nice touch.

However, there are a few things I noticed that I don’t love. The car on the right seems to drift, and the blinking headlights don’t quite match the usual rhythm. Additionally, while there’s a center stripe on the road, all the cars are driving in the same direction—no one is heading toward the red traffic light.

Bison Stroll on Broadway

I tried to use the preset to give the Bison character a defined name, but it didn’t work as intended. In this rendering, the Bison is depicted as a dog, which, while not perfect, is a significant improvement over some versions where the character is portrayed as a person. Despite this misstep, this rendering still ranks among my favorites.

The shallow depth of field is stunning, creating a sense of focus and immersion. The colors are vibrant and pop on the screen. I love how the dog meanders slightly, following its nose rather than sticking to a rigid path—it adds a touch of authenticity and personality. The fur and light reflections are incredibly detailed, and the dog’s shadow appears to track realistically throughout. Toward the end, as the dog crosses in front of a man, the cowboy-hat-wearing character performs a subtle do-si-do step. While a bit stereotypical, this detail is charming and adds character.

Now, the not-so-good. The rendering struggles with details in background elements. The double-wide sidewalk contrasts sharply with an unusually narrow road, creating an odd sense of scale. At the start of the video, one man on the right is wearing a onesie, inexplicably has three hands, and is missing a leg. Another man, dressed in a white shirt and dark jacket, has significant rendering issues with his left arm—it bends awkwardly at one point. Additionally, a woman in the background carrying a bag suddenly grows taller in a single frame, causing a noticeable jump.

Interestingly, I didn’t notice any major issues with the dog or the man in the cowboy hat, who are arguably the main characters in this scene. This leads me to believe that Sora is prioritizing the details of key subjects over those of background elements.

If that’s the case, it’s a logical approach. Most viewers focus on the main characters and actions, paying less attention to the background. By allocating its computational resources toward perfecting the shadows, reflections, and details of the central characters, the system is likely optimizing for the aspects viewers are most likely to notice.

Let’s look at one more before we move on to the not-so-good.

Nashville’s Neon Bison

This Bison was rendered using the “Balloon World” preset, which promises a glossy, balloon-like aesthetic with a high-key, fun vibe. It certainly delivered on that promise.

The Bison balloon is fun to watch, and it seems the crowd agrees. The reflections on its surface bounce perfectly as it "walks," creating an uncanny realism for something so whimsical. While I’ve never seen a walking balloon Bison in real life, I imagine the way the light reacts here is quite accurate. The shadows and lighting are, once again, impeccably executed, contributing to the overall charm.

As impressive as the Bison is, the background details leave something to be desired. Several signs have misspellings, such as “Broadway” rendered as “Bradway.” The balloons in the scene seem to float with no discernible attachment beyond their strings.

Early in the video, on the far left of the screen, a woman appears without facial details, and others around her have smeared, indistinct faces. The lines on the road also look odd—perhaps this is meant to be a pedestrian-only street? Meanwhile, the men on the right seem to be recording something with an oddly shaped phone—or playing with a wallet? The lack of clarity here detracts slightly from the scene.

This is another example of how the generator prioritizes key elements, like the main subject, over peripheral details. For most viewers, this trade-off likely works well, as attention naturally gravitates toward the focal point of the scene.

Now, the not-so-good…

Lou’s Broadway Stroll

Why is the Bison rendered on a greenscreen? In an AI-generated video, this feels unnecessary unless it’s meant to be ironic. Despite the visible green around the Bison, the shadows are impressively accurate. However, the slight misalignment of the Bison's shadow with the environment further emphasizes the greenscreen effect. This isn’t the only oddity, though—it almost feels like the AI is gaslighting me.

The sunset is overly vivid, bordering on artificial, and the two cowboy-hat-wearing figures on the left seem awkwardly inserted as if they don’t belong in the scene. Additionally, the overall video quality feels unnaturally sharp, resembling HDR footage more than traditional film. The mismatched road lines and pedestrians casually walking in the street only add to the sense that something is off.

Could the AI be intentionally producing an ironic “video” aesthetic?

The exaggerated vibrancy, the greenscreen effect, and the hyper-realistic sharpness might be deliberate choices to mimic or parody certain overproduced visual styles. This could be a nod to low-budget or experimental filmmaking, where visual imperfections are embraced as part of the charm. If that’s the case, the AI’s approach is clever but risks being misunderstood.

Studious Bison in Library

I did render a few videos beyond the Bison walking on Broadway. Take a look at this one: The Bison student studying in the library. Unfortunately, his glasses are awkwardly placed on his head, more like a poorly applied filter than an integrated part of his design. In the background, a girl at the table is either closing or opening her book—it’s hard to tell, but it is distracting. Meanwhile, the girl across from her holds a “magical” pen that appears and disappears in her hand.

But it is not all bad. The grain on the table is beautifully rendered, and the reflections highlight its polished surface. As usual, Sora does a great job handling light and shadows.

The recurring attention to shadows, light, and reflection makes me wonder about the specific training focus in these areas.

Other subtle touches also stand out: the swinging of the girl’s feet (on the far left) under the table adds a realistic charm, and the Bison’s fur is remarkably detailed. Remember when computer-generated characters struggled to render realistic hair? It wasn’t that long ago.

Campus Sunset Stroll

Last but not least are the walking errors. While I don’t mind that the Bison in this video resembles a stuffed animal rather than a realistic creature, I’m not sure why it’s walking backward. However, I love the symmetry. It adds an unexpected charm.

As in previous examples, the background characters present clear issues: missing hands, smeared faces, and an oddly animated package that seems to have a life of its own. The leaves on the tree also have a strangely digital appearance, which detracts from the overall realism.

Still, for all its shortcomings, Sora delivers impressive detail in other areas. The leaves scattered on the ground, the ivory on the building, and the windows are rich with detail.

It’s a fascinating blend of technical details and peculiar quirks. Did you see the girl on the left with the unicycle?

This wasn’t the only walking mishap I encountered today. I’ve compiled three videos that showcase some of the quirkiest movement issues, ranging from super-speed walkers and a backward stride to some peculiar—bouncing?

I like to call this: "An Ode to AI Movement Fail."

Vibrant Campus Life

Bison Takes Over Broadway

Bison Strolls Broadway

Sora demonstrates impressive potential—particularly in its handling of light, shadows, and fine details—but its inconsistencies reveal what we all expected, there is room for improvement.

We have to keep in mind that these technologies are still in their infancy.

The mix of impressive moments and humourous mistakes makes for a fascinating exploration. I’ll continue experimenting as I get more tokens. For now, I’m taking today’s findings with a blend of curiosity, a critical eye, and a healthy dose of laughter at those quirky bison moments!