Inspiration
The Shadows Weaver was born in a paradoxical space: that of documentary television and the realm of philosophical tales. Originally commissioned for the program PhantasIA (ARTE France's research workshop), it began as a documentary narrative before being adapted as a standalone piece. The Weaver subverts the conventions of storytelling, using them as a poetic structure to serve a chapter-based narrative that explores several themes related to Artificial Intelligence. It questions the gaze of & from the machine, the camera, and the viewer. It presents AI as both tool and character: a mirror that observes, measures, and archives, but fails to grasp the principle of singularity. Finally, it plays with the boundary between fiction and documentary, between staging and unveiling. It is the result of research into a new form of storytelling assisted by generative AI. The Weaver weaves images and meanings between the threads of history, the network of machines, and the synapses of our imaginations.
My core visual & conceptual inspiration: to twist the codes of the fairytale in order to reveal the normative patterns hidden in technologies that claim to be “neutral.” In that extend I used Public Domain illustration models I trained on Krea and mixed them with an other model trained on wacky selfies of myself. One could say, I am somehow the Weaver myself.
How we built it
The script was co-written with ChatGPT, following the structure of an initiatory and philosophical tale, questioning our relationship to singularity and uniqueness at the time of Artificial Intelligence. I created two separate custom image dataset using public domain anatomical engravings and illustrations, and trained them using Krea. I also trained a LoRA with wacky selfies of mine to add a bit of physicality in the character. I wanted to create an aesthetic between medical archive and gothic collage to support a dark fairy tale vibe.
Chat GPT helped with structuring the story board and generating prompts. Each video sequence was generated using Freepik prompt enhancer. Then, the videos were made with Kling 1.6 and 2.5, Hailuo, and Dreamina with the 10 frames interpolation feature.
The voice-over was generated with Runway, sound design using Eleven Labs. I composed two original scores using Udio: one in a baroque fairytale register (flute and harp), the other more spectral and ambient, echoing the PhantasIA interludes. Editing was done in Premiere Pro, with additional image edition work in Photoshop Firefly and Nano Banana. Final sound mix was completed by Etienne André.
Challenges we ran into
I am really happy with the aesthetic that came out from this experimentation. The hardest part was to choose images! But overall, the tools are getting better and better, so the challenges are fewer. I truly can't believe the improvement in just 6 months.
Accomplishments that we're proud of
My models. They are the core of the film and I am inspired to make more tales using this aesthetic.
What we learned
I learned to work within a broadcast framework while asserting a strong artistic identity. Along the way I discovered new tools like Dreamina, where the 10 frames animation feature was a game changer. I also was amazed with the prompt enhancer on Freepik that helped a lot. A fairly good surprise was also the Hailuo frame interpolation which made wonders especially on the book pages turning. In short: each AI project I jump into works into a timeframe where the tools are at the state where they are, and I need to adapt my expectations. But this time I feel like the tools have made unbelievable improvements. My workflow gets smoother and smoother and I can't wait to start working on my new film.
What's next for La Tisseuse d'Ombres - The Shadows Weaver
Now that the film is done, it will hopefully travel the world in Festivals. But The Weaver also inspired me with the artistic direction of a series of fairy tales I had in my idea-fridge. This will be my big project for 2026.
Built With
- adobe
- dreamina
- eleven-labs
- freepik
- hailuo
- kling
- krea
- photoshop
- premiere
- runway
- udio
Log in or sign up for Devpost to join the conversation.