Artificial intelligence is no longer just changing how films are marketed, edited, or distributed. It is beginning to reshape the entire production pipeline itself. From micro-drama production to filmmaking, where budget, resources, and creative iteration often determine whether a project can move forward, AI is becoming more than a tool. It is becoming part of the creative infrastructure. Here, we are referring specifically to internet-native micro-drama rather than traditionally defined films or television series, a format that began emerging around 2018 to 2019 and quickly gained mainstream popularity around 2020.

To better understand this shift, we invited industry expert Amy Kouxiao Zhang, one of the Co-Founders and the Chief Content Officer of Manna AI, to give us a real snapshot of how AI is entering the micro-drama production process, what has changed compared with traditional filmmaking workflows, and what challenges still remain as creators begin to rely on AI more deeply.

Amy's path to entertainment runs from the University of California, San Diego, where she studied Visual Arts Media with a minor in Computing Arts focused on Virtual Reality, to the American Film Institute Conservatory, one of the world's most prestigious film schools, where she earned her MFA and trained under Oscar-winning filmmakers. Her work as a filmmaker has been recognized at festivals including Cannes, Tribeca, the Student Academy Awards, Golden Horse, and Golden Rooster. After AFI, she worked inside Hollywood studios and distributors before moving into vertical mini-series, where she produced chart-topping titles.

Over the course of our conversation, Amy shared that she has spent the past three years working as both an independent producer following her graduate studies and as a creative producer on numerous mini-series projects. Her portfolio includes titles such as My Teenage Great-Grandma, Comeback of the Basketball King, and The Missing Kitchen King, all of which gained significant traction on platforms like TikTok, Pine Drama, and Netshort. Among them, My Teenage Great-Grandma achieved remarkable success, reaching more than 600 million views worldwide.

Today, Amy is embarking on a new chapter shaped by the rapid advancement of generative AI technologies such as Seedance. As the co-founder of Manna AI, A Silicon Valley AI-native tech & entertainment company, she is focused on reimagining how stories are created, experienced, and distributed in the AI era.

In order to achieve this vision and mission, she co-founded Dreem, Manna AI's flagship AI-native interactive storytelling platform, where creators can turn story ideas into watchable, playable, and remixable story worlds. Rather than simply generating AI videos faster, Dreem enables creators to build persistent characters, interactive narratives, and long-term IP ecosystems that extend beyond a single piece of content.

Before Generative AI: The Reality Behind Mini-Series Production

Amy described the pre-AI era of mini-series production as a highly collaborative and resource-intensive process. Like traditional film and television productions, mini-series projects required assembling an entire team before cameras could even begin rolling.

The process often started with open casting calls, where producers interviewed potential actors and evaluated their suitability for each role. According to Amy, casting decisions were influenced not only by an actor's performance abilities but also by factors such as their previous work and existing fan base, which could help drive audience engagement once the series was released. While acting talent remained important, the standards for mini-series production were often more flexible than those of traditional film and television, where performances and scenes might be much more repeatedly refined through multiple takes.

Despite being shorter in format, Micro-Drama productions still involved significant financial investment. Production budgets could easily reach 2 hundred thousand dollars, covering expenses such as cast compensation, studio rentals, production crews, equipment, and post-production work. Understanding these costs is particularly important when comparing traditional production methods with the new AI-driven trend that is reshaping the industry today. While Micro-Drama production costs are already significantly lower than those of traditional Hollywood productions, AI-driven production can reduce these costs even further.

Was artificial intelligence used during this pre-generative AI period? According to Amy, the answer is yes, but only in limited capacities. AI was primarily used as a supporting technology during post-production processes, such as assisting with green screen effects and other visual enhancements. Rather than fundamentally transforming the creative workflow, AI functioned as a behind-the-scenes tool within an otherwise entirely human-driven production process.

The Post-AI Production Era: Everyone Can Be a Director

According to Amy, the emergence of generative AI has fundamentally changed who gets to tell stories through film. Traditionally, bringing a script to life required substantial financial backing, industry connections, and years of production experience. Today, AI is lowering many of these barriers, opening the door for younger folks.

Amy noted that in the traditional Hollywood system, only a small fraction of scripts ever make it to production. Of the countless stories pitched to producers, she estimated that only around 1% are ultimately developed into films or series, while the remaining 99% never reach broader audiences. As a result, many compelling ideas and diverse perspectives go untold.

With advances in AI-generated video technology, however, creators no longer need access to Hollywood funding or extensive filmmaking experience to visualize their ideas. Individuals with a compelling story can now transform scripts into mini-movie clips using AI-powered tools, significantly reducing the barriers to entry.

Amy also highlighted how AI is changing the economics of production. Large-scale scenes that would have traditionally required enormous budgets, including but not limited to visually complex sequences similar to those seen in major superhero franchises such as Marvel films, can now be created at a fraction of the cost. What was once accessible only to studios with substantial resources is increasingly within reach of independent creators and small production teams.

Despite these advances, Amy emphasized that current AI technologies still face important limitations. One of the most significant challenges is maintaining character consistency throughout a production. Unlike traditional filmmaking, where actors naturally preserve continuity across scenes, AI-generated characters often struggle to retain consistent appearances and behaviors over time. While the technology has made remarkable progress, achieving the same level of coherence and emotional authenticity as real-time human performers remains an ongoing challenge.

The Existing Challenges of AI-Native Production

Despite the remarkable advances in AI-generated video technology, Amy emphasized that today's AI production workflows remain far from fully automated. One of the biggest challenges lies in maintaining consistency throughout a project. Amy explained that her team often breaks a script into multiple 15~30-second segments: "What we do is split the script into many 15~30-second clips and use editing tools to combine all of the AI-generated videos into one complete story."

However, creating a coherent final product remains difficult. Maintaining consistency in backgrounds, environments, and character appearances across multiple generated clips is still one of the industry's most persistent challenges. To address this issue, producers frequently provide AI systems with existing character images and ask them to generate videos based on those references, aiming to create smoother transitions between scenes and preserve visual continuity.

When asked what happens when AI hallucinates, Amy laughed and responded, "It happens all the time." According to Amy, there is no perfect solution other than patience, iteration, and careful guidance. Creators continuously refine prompts and regenerate scenes until the desired results are achieved. "Better prompts almost always lead to better results," she noted.

Cost efficiency is often highlighted as one of AI production's greatest advantages, but AI-generated content is not entirely inexpensive. Amy estimated that producing approximately 100 minutes of AI-generated mini-series content currently costs around $20,000, with expenses largely determined by token consumption.

Generation speed can also vary significantly. Under ideal circumstances, AI video generation may take only 30 minutes to an hour. In more challenging cases requiring multiple iterations and refinements, the process can extend beyond a full day. Perhaps one of the most relatable aspects of AI production, Amy joked, is that the technology sometimes seems to perform best when creators step away from the screen. "There have been times when we became frustrated, took a lunch break, and came back to find that the AI had suddenly generated exactly what we were hoping for," she said.

These challenges are also part of what Dreem is trying to address. Its Creator Studio helps AI filmmakers simplify the production process, particularly around recurring challenges such as maintaining character and scene consistency across generated content. Meanwhile, Dreem’s mobile app focuses on the other side of the equation: distribution, giving UGC creators a place to share their stories, reach broader audiences, and potentially earn from their work. Together, the two products reflect Dreem’s broader goal of making AI-native storytelling more accessible: not only to create, but also to be seen.

While AI has dramatically lowered barriers to entry for filmmakers, Amy's experience suggests that successful AI-native production still requires a combination of creative vision, technical expertise, prompt engineering, and, perhaps most importantly, patience.