The Gap Between Demo and Reality
The headlines are full of buzz about text-to-video models like Google’s Veo and OpenAI’s Sora, which can generate stunningly realistic video clips from simple text prompts. These tools, often part of larger AI families like Gemini, promise to revolutionise
content creation. The demos are breathtaking, showcasing cinematic shots and complex scenes generated in seconds. However, for creative professionals, the current reality is far from this seamless vision. The most significant limitation is the length; most tools are capped at generating clips that are only a few seconds long, often around eight seconds. This makes creating any form of long-form content impossible for now. Beyond the time limit, there are severe constraints on iteration and refinement, with daily generation caps and credit systems that restrict experimentation, a core part of any creative process.
Losing the Director's Chair
The core of the problem is a fundamental lack of creative control. Filmmakers, animators, and designers rely on precision. They need to direct a character's specific glance, control the subtle shift of light, and maintain shot-to-shot consistency. Current AI video tools struggle with this. A major issue is character persistence; generating the same character across different prompts or scenes is nearly impossible, as the AI creates a new interpretation each time. This breaks the basic continuity required for any narrative. Furthermore, the output can often have an uncanny, artificial feel, with strange physics or lighting that a creator cannot easily fix. While some artists embrace these AI “accidents” as part of the art form, for commercial projects requiring brand consistency and narrative coherence, this unpredictability is a major roadblock. The tools offer endless possibilities but little purposeful control.
The Urgent Need for Ethical Review
Beyond the practical limitations, there's an urgent need for robust ethical review processes. These AI models are trained on vast datasets scraped from the internet, which means they can inherit and amplify existing societal biases related to gender, race, and profession. An AI might default to stereotypical depictions if not explicitly guided, creating problematic content. This necessitates a human-in-the-loop approach to review outputs before they are published. The risk of creating deepfakes, spreading misinformation, and infringing on copyright are also significant. Tech companies are beginning to implement safeguards like watermarks and policies against creating likenesses of real people without consent. However, these are often reactive measures. A proactive ethical framework must be built into the workflow, demanding transparency about AI use and ensuring content is fact-checked and culturally sensitive.
A Tool for Co-Creation, Not Replacement
The argument is not to abandon these powerful new technologies. Instead, we must frame them as what they currently are: powerful creative assistants, not autonomous directors. Their strength lies in augmenting human creativity, not replacing it. They can be invaluable for brainstorming, creating storyboards, or generating abstract B-roll footage. Some emerging platforms are even building workflows around an AI agent that helps organise ideas on a canvas before generation, attempting to bridge the gap between a simple prompt and a finished product. The future of creative work will likely be a hybrid model, where artists use AI for specific tasks while retaining final creative judgment and control. This approach combines the efficiency of AI with the authenticity and intention that only a human creator can provide. To get there, developers must prioritize building features that empower creators with granular control rather than just chasing more realistic output.
