Last week, CNN published a report about the explosive growth of AI-generated microdramas in China. More than 120,000 microdramas were released in the first three months of 2026, and around 95% of them were generated using AI.
A year or two ago, most AI video discussions were about whether a model could keep a character’s face consistent for ten seconds. Now entire production pipelines are being built around these models, and creators are publishing serialized stories at a volume that would have been impossible with traditional production.
I’m pretty sure anyone who spends enough time on TikTok, Instagram, or Facebook has already watched at least one of these AI dramas.
Some are obviously generated, especially when the motion gets complicated or objects start behaving strangely. Others are convincing enough that I sometimes only notice after a character changes slightly between cuts or a background detail suddenly disappears.
What’s even more interesting is how these productions are set up and how the final video series are actually being made.
There are currently a handful of video generation harnesses and web platforms that let you create full-length AI films with very little effort.
For example, Topview’s Drama Studio and Film Studio are trying to organize the messy production work around the models, from story development and character creation to storyboarding, scene planning, generation, and editing.
For anyone curious about how these AI dramas are being produced, the process is a lot more structured than repeatedly typing prompts into a text-to-video box.
Go Short or Long Form AI Film
Topview currently offers two filmmaking products that serve different kinds of creators.
Drama Studio is designed for microdramas and episodic stories. I can start with an idea, outline, script, or longer piece of source material, then let AI agents help break it down into scenes, characters, environments, storyboards, and video.
This is what a sample video series built with Drama Studio looks like:
As you can see above, multiple scenes are getting stitched together to form a long and multi-scene video. The result is a few minutes of video clip that could be an episode of a series.
Here’s what the final video looks like:
Film Studio is for projects where I want much more control over the actual shot. It gives me tools for character performance, camera direction, scene composition, blocking, and visual refinement, which makes it closer to a filmmaking workspace than an automated series generator.
As you can see above, the dashboard gives you tons of freedom and control over each scene in the sci-fi film. You have a chat box on the right side, which lets you delegate any work to an AI agent instead of you manually doing all the work.
Here’s what the final video looks like:
Pretty cool, right?
Both Drama Studio and Film Studio are powered by Seedance 2.5 model, so you can expect high-quality and realistic video sequences every single time.
I would not overthink the choice between them. If the main problem is producing several connected episodes efficiently, Drama Studio makes more sense. If the project depends on precise framing, performances, or how one shot connects to another, Film Studio is the better fit.
I sometimes create short AI films just for fun and send them to my mom afterward. Most of them are little fantasy or science-fiction ideas that would never justify renting equipment, finding actors, or learning a complicated 3D animation pipeline.
Being able to make those scenes at home is fun on its own, but the same tools become much more interesting when I imagine what an independent filmmaker with a finished script could do with them.
Topview’s Drama Studio
I previously wrote a more detailed guide to Drama Studio, so I will keep the walkthrough shorter here. You can read it below.
The Fastest Way to Create an AI Series with a One-Person Drama StudioYou don’t need a crew, a camera, or a budget to make an episodic drama anymore. You need a story and an afternoon.The thing that you need to know about this tool is that it treats a drama as a production instead of a collection of separate generations. I give it a story, then the agent helps organize the pieces needed to turn that story into an episode or series.
Some of its key features include:
- AI-assisted script development for turning an idea, outline, script, or longer story into scenes and dialogue.
- Reusable characters, locations, and props that stay connected to the same project.
- Storyboard generation before committing to more expensive video renders.
- Shot-level editing and regeneration when only part of an episode needs to change.
- Multi-episode continuity for stories that reuse the same cast and environments.
When I tested Drama Studio earlier this year, I used it to create a fantasy confrontation between a human paladin and an orc warrior. The original idea was pretty simple, but the system helped expand it into something closer to an actual sequence by creating characters, establishing the environment, breaking the action into scenes, and preparing the visual direction for each shot.
Without a system like this, I would normally handle all of those steps manually. I would generate a character image, save it somewhere, create a second character, keep track of the references, write prompts for every new shot, and repeatedly remind the video model what everyone is supposed to look like. Once the story reaches several scenes, keeping all of that consistent becomes more annoying than writing the story itself.
Drama Studio keeps those assets attached to the same project, so I can reuse the characters and environments instead of rebuilding the context every time.
Here’s a sample project board I have with one of my mini film series.
The storyboard stage also gives me a cheap place to make mistakes. If the orc is standing on the wrong side of the frame or the paladin suddenly has different armor, I would rather notice that in a still image than after generating the final video.
This is why I think Drama Studio fits microdramas particularly well. Romance series, fantasy serials, short horror stories, anime-inspired episodes, or vertical dramas for mobile platforms all reuse characters and locations heavily.
Producing five or ten episodes a week is less about creating one spectacular shot and more about keeping the production moving without losing consistency.
I still would not publish everything the agent writes or generates without reviewing it. Dialogue usually benefits from a human rewrite, pacing needs judgment, and some scenes simply need several attempts before they work. But I would rather spend my time fixing those creative problems than manually rebuilding the same production setup for every episode.
Topview’s Film Studio
Film Studio goes further in the opposite direction. Instead of trying to automate as much as possible, it gives creators more control over what happens inside an individual shot.
That approach is closer to how I think serious AI filmmaking will develop.
Prompting works well when I am flexible about the result. If I write “a woman waits nervously in an empty train station,” a good video model can make something visually impressive from that sentence.
But filmmaking becomes more specific once I care about how nervous she looks, whether she avoids eye contact, where she stands relative to the camera, or how slowly the camera moves toward her.
Those details are difficult to control through text alone.
Film Studio introduces controls that separate some of those decisions from the prompt itself. Character expressions, camera placement, scene composition, and blocking can be treated more deliberately instead of being bundled into one paragraph of instructions and left to the model to interpret.
Here’s an example mini AI film I am currently working on.
As you can see, every character can be controlled by a series of instructions. You decide her story, her looks, her background, and personality.
This is a sample snippet of the mini AI film:
This information gives the AI agent an idea of how the character should look and express in every scene.
AI models can generate beautiful faces now, but subtle acting is still inconsistent. A character might react too strongly, smile at the wrong moment, or exaggerate an emotion that should have been barely visible.
Giving creators a visual way to adjust the emotional performance makes more sense than rewriting “slightly worried, but trying not to show it” five different ways and hoping the next generation gets closer.
The same applies to camera work. A camera angle is not just decoration. Moving closer to a character can make a conversation feel uncomfortable.
Shooting from behind someone can hide information. Keeping the camera wide for a few extra seconds can change the timing of a reveal. When these choices are left entirely to the model, the result might look cinematic without actually saying what I wanted the scene to say.
