StoReel, a Chinese AI-native microdrama platform, on Tuesday unveiled a new tool called Canvas that lets anyone create their own show. TheWrap got early access to the tool and decided to make a microseries, authored by an entertainment reporter. Why not?
After four days and about $150 worth of digital credits to access the tool, we created four minute-long episodes designed as a take on the popular hockey romance series “Off Campus.”
You can check it out below. The “actors,” the sets and the script were all generated by AI.
What’s clear is why these microseries companies are so keen to use AI, and why a flood of AI-generated content is already widely available and thriving, especially overseas. Using Canvas to finish the show would’ve likely taken a few thousand dollars more, while an average live-action microseries costs $200,000 to produce. The lower cost and short time means these companies can flood the market with more series at a time when the appetite for the medium has been insatiable.
But as we demonstrated with our attempt at a microseries, just having access to these tools isn’t enough. There’s the “uncanny valley” aspect with acting that looks stiff and not quite human. At-times odd camera moves and the occasional awkward visual force a double take — weren’t those two characters just sitting on the other side of the couch? Where’d the wall go? But all that aside, it’s clear that someone with serious of vision and talent could use this tool to make something worthwhile.
“If the tech is already so advanced, what’s really going to be competitive is everybody’s taste,” StoReel co-founder Angela Yu told TheWrap. “It’s not going to be that AI is going to take over creation, but people who actually have very high creation powers are going to be the ones that succeed.”
Because if the story is good, users don’t seem to mind.
“We saw that the metrics of engagement retention for AI-generated content based on human screens … is really similar to what we do with live action,” Bogdan Nesvit, CEO of Ukrainian vertical platform maker Holywater, revealed earlier this month at a microdrama conference in Los Angeles.
The process
Most AI-generated microdramas from the leading companies – including Holywater’s MyMuse, StoReel, DashVerse and Shortical – typically employ an entire team to write the script, but we enlisted Claude and ChatGPT for assistance.
After the immense success of hockey romance series “Off Campus,” we decided to create our own spoof version of the highly anticipated Allie Hayes and Dean Di Laurentis love story set to take place in Season 2.
We found that the AI-generated series made with StoReel can cost as little as $50 per minute. We created four episodes of our “Off Campus” spoof for over 11,000 StoReel credits, which amounts to roughly $150. In comparison, a typical live-action microseries costs roughly $2,000 per minute of content.
As we are not screenwriters by trade – and did not have the time to write a 30-page script ourselves – we asked ChatGPT to generate a 30-page / 30-minute microdrama. Here’s the prompt we used:
I want to write a 30 page/30 minute microdrama script based on Elle Kennedy’s “The Score.” I want the script to follow Allie Hayes and Dean Di Laurentis’ love story and feel like what could be Off Campus Season 2. Can you divide the script up into 1 to 2 minute episodes like typical microdramas scripts with cliffhangers that end each episode? The characters and plot should be based off of Elle Kennedy’s book but also the tone of the Amazon Prime series. Please generate a 30 page/30 episode script titled Off Campus Season 2.
We wanted to test if the AI generation tools would reference the source material directly and if they would be able to write in the style of microseries’ notorious cliffhangers and romantic twists.
ChatGPT did not provide a script, stating that the tool “cannot generate a scene-by-scene script or screenplay that closely recreates a copyrighted book’s characters, plot,and dialogue.”
Instead the tool spit out an adaptation outline.
We then moved to Claude, giving the tool the same prompt and within five minutes the tool produced a 50-page script based on the hockey romance novel.
From there, we fed the script into StoReel’s Canvas tool, which has a one-click process that broke down the script into characters and settings. We were then able to edit physical attributes or details about each character to further tailor the specific features before generating spoof versions of the “Off Campus” stars.
We attempted to tell the tool to make each character look like their corresponding Prime Video counterparts directly, but the technology did not respond to that request. Creating a character in someone’s likeness was not allowed, a potential barrier to guard against copyright violations.
StoReel generated detailed physical elements, including descriptors like a “smooth and plump perioral region in the face” or “almond-shaped doe-eyed green irises.”
While Allie and Dean came out close enough to their original characters, Garrett Graham and some of the other characters looked significantly different than the series, so we had to pay to regenerate them.
Generating the characters and settings cost 288 credits. This amount was included in the free 300 credits each StoReel Canvas user receives. But each edit required further video generation and more credits, and we found ourselves going back to StoReel for additional tokens, returning several times to fully generate four episodes of the microseries.
Creating the actual video episodes also took time. The entire 30-episode storyboard was divided into 146 clips. Each storyboard clip amounted to about 10 to 15 seconds of content.
StoReel created an intricate storyboard for each clip, providing detailed stage directions for how each AI actor should behave. Each prompt was between 300 and 400 words, detailing how the actors should move, where the camera should go and where the lines fit in between the action. Once we paid for generation, it took roughly 10 minutes to produce a 10-second clip.
The result
While the video generated was of realistic quality — nobody’s hand had six fingers and the requisite shirtless male character scene had appropriately glistening abs — the acting felt wooden at times. (Although even human actors sometimes deliver wooden scenes.) The actors often took unnatural breaks between lines or stared more intently than in a typical human interaction. Some characters delivered lines that were devoid of emotion.
We did not make edits to the stage directions or camera cuts provided by StoReel.
We also noticed that in video generation, StoReel changed the author of the featured book from Steinbeck to Steinberg. It was unclear if this was a hallucination or an intentional change to avoid a copyright infringement from the author. The system did not seem to register that the script was taken directly from the plot of Kennedy’s “The Score.”
We generated four episodes: the opening scene where they meet, one of the first hockey scenes where Dean is ruminating on the ice, the morning after where the couple decides their feelings are deeper than expected, and a climatic scene where Allie and Dean reveal their true feelings for one another.
Professional user-generated content
Launched this week, StoReel wants to encourage people to use Canvas to create their own microseries that could actually end up on StoReel’s streaming platform. The company deems this content professional user-generated content, or PUGC.
Anyone can use the company’s AI engine StoReel Canvas to create their own microseries. The platform also offers paid opportunities where independent creators can contribute to existing titles that StoReel already has in progress.
StoReel shells out 1,000 Chinese Yuan or about $150 per minute of content creation for series they designate as PUGC.
These videos are available to stream on the company’s direct-to-consumer viewing platform StoReel, which is available to download on Apple and Google devices.
“About 20% of content comes from in-house and 80% is from the PUGC creator ecosystem, but by the end of the year, our goal is to do 100 shows per month,” Yu said. “We’re just trying to scale this content as fast as possible, while ensuring quality.”
Think of it less like a standard microseries platform with highly produced content, and more like YouTube. It’s about quantity rather than quality.
“We’re looking to build a YouTube, not a HBO or a Disney or Netflix, so we want the content itself to scale, not to scale it linearly like a traditional studio,” she added.
For its own content, StoReel said it spends 10 days to a month on the scripting process and a month in the production. While the production process is still much quicker and cost efficient than a live-action shoot, the company argues the productions still require artistic skill.
Initially, the co-founders wanted to just create a traditional microseries app, but they saw the steep competition and how much other companies were spending on user acquisition, and it was not sustainable as a startup.
“In January 2025 we saw the [AI] content was just terrible, it was unwatchable, but in September, when Veo 3 was launched, we saw that there was a huge quality jump,” Yu said. “Now the quality of AI short dramas are really becoming commercially viable.”
One of the company’s AI-produced titles “OMG! My Snowwhite Is a Man” recorded a cost per install of $4.63, below the typical $8 to $10 range the company reports for its traditional live-action dramas. The company also reported that among its AI-native catalog, which includes more than 80 original series, average Day-7 retention runs 22% higher than comparable live-action titles.
StoReel’s main revenue model is subscription payment. In the North American market, the price of subscriptions starts at $19.99 per week with an annual membership costing $269. The company has also brought its own costs down by creating a user-generated content hub to create more volume.
Yu explained that in China there is much more acceptance around AI content. She noted that as opposed to the U.S., Chinese people see it as an opportunity that they want to embrace. She warned that Americans could be falling behind by boycotting the technology altogether.
“[Chinese people] want to embrace AI because they think it’s going to be part of our lives forever,” she said. “It’s not that it’s going to take everybody’s jobs, it’s just going to take people’s jobs who don’t know how to use AI.”
For a relatively inexperienced AI user like us, this project showed that not just anyone — or any script, for that matter — can lead to something worth watching.
Like Yu said, it takes taste.
