{"slug": "what-17-minutes-of-ai-filmmaking-really-cost", "title": "What 17 Minutes of AI Filmmaking Really Cost", "summary": "Filmmaker Javi López released The Backrooms Nemoris: It Eats Light, a 17-minute AI-generated short film that took six months to produce, required roughly 12 hours of generated footage from approximately 7,000 generations, and cost an estimated $20,000 in generation fees. López spent about 60% of production time on sound design in DaVinci Resolve, and some handheld-shot prompts needed around 50 generations to yield a usable result. The project marks a push past the eight-minute ceiling of prior AI short films toward sustained narrative.", "body_md": "## A 17-Minute Film, Six Months in the Making\n\nFilmmaker Javi López recently launched **The Backrooms Nemoris: It Eats Light**, a 17-minute AI-generated short film. This project significantly pushes the boundaries of AI filmmaking duration, as the host of Theoretically Media noted his own previous AI film ceiling was just eight minutes. López aimed to move beyond short-form experiments into sustained narrative.\n\nLópez chose **The The Backrooms** as his narrative canvas, tapping into its rich internet mythology. Originating from a single 2019 image on 4chan, The The Backrooms evolved into a collaborative, open-ended liminal space, famously popularized by Kane Parsons’ influential 2022 YouTube series. This shared lore provides an ideal sandbox for creative reinterpretations.\n\nWhat began as an anticipated one- to two-month production ballooned into a **six-month odyssey**. The 17-minute runtime isn't the product of a simple prompt-and-render; it represents a concentrated, solo effort, testing the true endurance and cost of current AI filmmaking tools in a feature-like format.\n\n## The Camera Was AI. The Director Wasn’t.\n\nJavi López didn’t delegate his vision to an agentic AI workflow. He retained precise, human control, structuring prompts around granular details: location, specific **handheld camera instructions**, action, dialogue, and even a \"no music\" directive. This wasn't an AI directing; it was López using AI as an unpredictable, but ultimately subservient, camera.\n\nConsider the sheer scale of output required for 17 finished minutes: López generated roughly **12 hours of footage**. Research estimates place the total generations at approximately 7,000, incurring an estimated $20,000 in generation costs. These are estimates, of course, but they offer a stark look at the compute required.\n\nLong, continuous handheld shots proved particularly costly. López revealed some prompts demanded around **50 generations** to yield a usable result. This contrasts sharply with traditional filmmaking, where low-resolution tests or drafts can quickly weed out bad ideas before expensive refinement. In AI filmmaking, every generation costs, and achieving continuity from an erratic digital lens demands a heavy toll.\n\n## The Scariest Work Happened After Generation\n\nGenerating images proved only one phase of López's Herculean effort. The real, often terrifying, work began *after* generation: a brutal curation process of selecting, pacing, and assembling thousands of unpredictable clips into a coherent 17-minute viewing experience. Each generated clip arrived with random artifacts, inconsistent motion, and a distinct lack of narrative flow, demanding a meticulous, human-driven hand.\n\nSound emerged as the unsung hero, the hidden engine of immersion. López dedicated roughly 60% of his entire production time to **sound design** within DaVinci Resolve. He layered generated audio with custom recordings and library effects, sometimes combining over a dozen tracks to craft a single scene. This intricate sonic tapestry compensated for visual uncanny valleys and stitched together disconnected AI-generated imagery.\n\nThe practical lesson is stark: sound is a production pillar from the outset, not a last-minute polish. Even when visuals felt alien or disjointed, expertly crafted audio provided crucial elements of place, tension, and **continuity**. This holistic approach to filmmaking, where the non-visual elements carry significant narrative weight, is critical for pushing the boundaries of AI-assisted storytelling, as Javi López and the team at [Magnific AI AI - Official Website](https://MagnificAI.ai) continue to explore.\n\nEnjoying this? Get one like it in your inbox each morning.\n\none email a day · unsubscribe in two clicks · no third-party tracking\n\n## Don’t Build the Next One Alone\n\nJavi López’s clearest lesson from *The The Backrooms Nemoris: It Eats Light* was the crushing weight of solo production. Six months of work, 17 minutes of film, and an estimated $20,000 in compute costs led to isolation. For the next installment of *The Backrooms Nemoris*, López plans to collaborate, specifically seeking:\n\n- Professional sound designers\n- Voice actors\n- Color graders\n\nHis experience offers a scalable workflow for others: favor hard cuts over continuous handheld shots to manage unpredictable AI outputs. Test concepts with draft generations, reserving high-quality, compute-intensive generation for shots that survive rigorous review. This iterative process conserves resources and streamlines editorial decisions.\n\nAI tools undeniably expand the ambition of a small creator, enabling a single person to generate 12 hours of footage from 7,000 generations. Yet, they do not eliminate the need for human judgment or, crucially, teamwork. The next leap in **long-form AI film** will combine generative power with specialist **craft**, transforming isolation into collaboration.\n\n## Frequently Asked Questions\n\n### How long did it take Javi López to make the 17-minute film?\n\nThe project took about six months, much longer than López initially expected.\n\n### How much AI-generated footage did the film require?\n\nLópez said the generations added up to roughly 12 hours of footage for a 17-minute final film.\n\n### Why was sound design so important to the film?\n\nLayered sound helped create atmosphere and continuity where generated visuals and their native audio could feel disconnected.\n\n### What advice did López offer to creators making AI films?\n\nUse lower-resolution draft generations to test ideas, avoid difficult continuous handheld shots, and consider building a team.", "url": "https://wpnews.pro/news/what-17-minutes-of-ai-filmmaking-really-cost", "canonical_source": "https://www.stork.ai/blog/what-17-minutes-of-ai-filmmaking-really-cost", "published_at": "2026-10-08 17:52:24+00:00", "updated_at": "2026-10-08 18:17:38.097840+00:00", "lang": "en", "topics": ["generative-ai", "artificial-intelligence", "ai-products"], "entities": ["Javi López", "The Backrooms Nemoris: It Eats Light", "The Backrooms", "Kane Parsons", "DaVinci Resolve", "Magnific AI", "Theoretically Media", "4chan"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/what-17-minutes-of-ai-filmmaking-really-cost", "markdown": "https://wpnews.pro/news/what-17-minutes-of-ai-filmmaking-really-cost.md", "text": "https://wpnews.pro/news/what-17-minutes-of-ai-filmmaking-really-cost.txt", "jsonld": "https://wpnews.pro/news/what-17-minutes-of-ai-filmmaking-really-cost.jsonld"}}