By Ian Dean
For centuries, Miguel de Cervantes’ literary masterpiece Don Quixote has stood not only as the foundational text of modern Western literature but also as a notorious graveyard for cinematic ambition. From Orson Welles’ decades-long, unfinished endeavor to Terry Gilliam’s legendary, disaster-plagued productions—famously chronicled in the documentary Lost in La Mancha—the epic tale of the delusional knight-errant has routinely defeated generations of filmmakers. The sheer scale, logistical nightmares, and tonal complexities required to bring La Mancha to life have historically made the project an industry-wide insurance risk.
Until now.
In Madrid on September 23, directors Guillermo Miranda and Javier de la Chica achieved what Hollywood heavyweights could not: they premiered a completed, feature-scale adaptation of the iconic windmill episode titled La Triste Figura (The Sorrowful Figure).
The catch? The comedy-adventure was constructed primarily using generative artificial intelligence tools, co-produced by and built upon the Magnific platform. Far from a silent, cold demonstration of tech capability, the film played to a packed art-house cinema in Madrid, eliciting genuine laughter, resonant dialogue appreciation, and a deeply human emotional response from its audience.
Main Facts: A New Frontier for an Old Knight
La Triste Figura marks a watershed moment in the intersection of artificial intelligence and narrative filmmaking. Co-directed by Guillermo Miranda and Javier de la Chica, the project bypasses traditional multi-million-dollar CGI pipelines, leveraging a suite of generative AI tools—chiefly Seedance 2.5 for video generation and Seedream 5 Pro for character and asset creation—to construct its approximately 120-shot runtime.
Unlike prior viral AI video shorts that lean heavily on slick, style-over-substance visual trickery, La Triste Figura grounds itself in rigorous pre-production design. The filmmakers did not merely prompt their way through the project; they built detailed turnarounds, expression sheets, costume tests, combat poses for the story’s "giants," and master proportion references.
Yet, the production consciously resisted total technological automation. Rejecting early AI-generated voices for failing to capture historical archaisms and dramatic weight, the directors hired professional actors Arturo Liberos and Juan Rabadán to voice Don Quixote and Sancho Panza. Furthermore, they integrated traditional human craft by bringing on storyboard artist and assistant director Víctor Ponce, alongside screenwriter Álex Rey.
The resulting 120 shots required roughly 1,200 total generations, pushing the boundaries of what a small, distributed team can achieve across disparate geographical locations.
Chronology of an AI-Driven Production
The journey toward La Triste Figura was years in the making, reflecting both the evolution of generative tools and the filmmakers’ persistent creative drive.
- 2019: Javier de la Chica begins experimenting with early-generation AI tools, recognizing the technology’s potential as a foundational medium rather than a mere novelty.
- Pre-Production Phase: The directors collaborate remotely across different Spanish cities, establishing a shared digital workspace. Recognizing the technical pitfalls of AI—such as Quixote’s asymmetrical design which prevented simple shot-flipping—they meticulously build character assets and visual references in Seedream 5 Pro.
- Human Integration: Writer Álex Rey comes aboard to tune the comedic timing of the dialogue, while storyboard artist Víctor Ponce hand-draws foundational framing, acting effectively as an assistant director. Voice actors Arturo Liberos and Juan Rabadán record the central performances, discarding synthetic vocal tracks that lacked narrative gravity.
- Production and Iteration: Utilizing Seedance 2.5, the team generates video shots based on Spanish-language prompts and visual references. Directors review and iterate in real-time, treating the AI pipeline like a "multiplayer game."
- September 23: La Triste Figura holds its world premiere at a boutique art cinema in Madrid, giving creators and critics their first look at audience reception for an AI-assisted narrative film of this scale.
Supporting Data and Technical Metrics
The production of La Triste Figura offers a fascinating window into the resource allocation, data metrics, and structural realities of AI-assisted filmmaking:
- Team Size: While official film notes list a core team of 10 people, co-director Miranda notes that headcounts fluctuated, scaling up to roughly 30 individuals at peak production intervals.
- Generation Efficiency: Crafting the final cut required approximately 1,200 separate generations to yield the roughly 120 shots featured in the movie.
- Runtime Expansion: Technical roadblocks occasionally dictated narrative length. Resolving complex generative errors—such as warping windmill blades or shifting cloak alignments—led the directors to expand certain sequences, adding roughly a minute and a half to the final runtime.
- The Dialogue Challenge: While AI models excel at generating high-octane action sequences and special effects spectacles, the filmmakers identified dialogue scenes as the ultimate test. A central five-minute sequence featuring four characters conversing against a backdrop of turning windmills consumed nearly half the film’s runtime and represented the production’s most complex technical hurdle.
Official Responses and Industry Perspectives
The reception within the creative community has been marked by a mixture of fascination, pragmatism, and acute anxiety. Miranda and de la Chica have been remarkably transparent about the industry polarization surrounding their work.
"We are trying to find the grey area, open the conversation," Miranda explains, noting that some industry peers have expressed outright hostility toward their use of generative tools, while others watch with quiet, cautious interest. "There are many shadows around the technology that we are not hiding," he adds.
Addressing common Silicon Valley narratives claiming that generative AI democratizes filmmaking by making it universally cheap and accessible, de la Chica pushes back. While La Triste Figura was produced at a fraction of a traditional studio budget, it still required substantial capital, institutional backing via Magnific, and a skilled team. "It is not yet a full democracy of the medium," de la Chica remarks.
Furthermore, the directors emphasize that AI workflows are not a universal panacea. Miranda, who also works professionally in medical animation, notes that generative tools have no place in fields where absolute physical and scientific accuracy is non-negotiable.
When discussing employment and studio structures, de la Chica predicts a contraction in traditional studio headcounts. "Studios with hundreds of artists will employ fewer people," he forecasts, though he hopes the lowered barrier to entry will foster a proliferation of smaller, independent boutique studios.
Implications for the Future of Cinema
La Triste Figura forces a radical rethinking of authorship, labor, and workflow in the digital arts. In traditional CG animation pipelines—such as those found in studios housing hundreds of artists—creatives frequently spend years isolated within specific departments (lighting, rigging, or rendering), only discovering the thematic thrust of the final film at its premiere.
By contrast, Miranda and de la Chica operated in a fluid, "multiplayer" environment where directors watched generations evolve in real-time and artists exerted macro-level creative control. This shift changes the psychological relationship between artist and medium. Animators accustomed to calculating motion to the millimeter had to "unlearn" rigid instincts, embracing an element of digital chaos that the directors playfully leveraged during the editing process.
Yet, the film’s ultimate triumph does not lie in its tech stack, but in its human anchoring. The deployment of hand-drawn storyboards, human actors for voice performance, and physical reference acting—such as de la Chica acting out Quixote’s sword-propulsion on the floor—demonstrates that generative AI functions best not as a creator, but as a catalyst.
As the dust settles on Madrid’s art-house premiere, La Triste Figura remains an experiment. Whether audiences will broadly accept AI-assisted features remains an open question. But by finally breaking the curse that stymied Welles and Gilliam, Miranda and de la Chica have proven that the windmill has finally been conquered—even if the giants were made of algorithms.
