We do not want to build one more “type a line, get a line” tool. In that model, the human still has to supply the real inspiration, and the machine only polishes the surface.
What we are building is an audience-emotion engine that can produce. It reads the emotional pressure carried by a crowd, then lets worlds, characters, plots and design directions keep growing. Human beings still judge and direct, but the story base no longer starts from a blank page every time. We are not designing a product. We are keeping a symbolic life form alive. This is Project Thalassa.
Traditional IP development is often a workshop cycle: design a character, write a story, launch it, spend it, and then begin again. We want a different form. The world, the characters and the audience feedback should all sediment into one base, so that the same base can grow new content again and again.
That is why the second slide uses a biological image. Thalassa is not mythology on a slide. It is closer to a mycelium network: living in social soil, absorbing audience emotion, spreading slowly, connecting worlds, roles and stories into one growing asset.
But today’s AI has three walls in front of it.
The first wall is emotion. Many tools say they understand feelings, but they mostly attach labels. In a UK context, think of a seaside pier, a football chant, a Christmas advert, a children’s book remembered by parents and grandparents. A tag can say “nostalgia” or “positive sentiment,” but it cannot tell you why that image still travels across generations. A few labels cannot contain the cultural memory that makes a story feel alive.
The second wall is sameness. The model becomes correct and smooth, but the content feels like the safe average. The third wall is tool dependence. It translates prompts. It grows only as far as the human input already reached.
So what are we actually making? In one sentence: a near-automated content-production line built around an audience-emotion engine. It does not help someone decorate a prompt. It starts by understanding an audience, then grows all the way into story, character, storyboard and sample film.
First, it listens. We do not ask only what one writer thinks. We listen to public, licensed and group-level material. Take a British seaside town. On the surface, it is a pier, a windy promenade, chips in paper, arcade lights, a coach trip in the rain. Underneath, it carries childhood memory, local pride, seasonal decline, small-business hope and the question of whether a place can still feel loved. That is safe material for story development because it is cultural and collective, not a private psychological profile.
Second, the system lets emotion, words, images and world settings bind into a story embryo. Third, it translates that embryo into characters, plots, boards and design direction. The point is simple: audience emotion comes in; production-facing story assets come out.
Technically, the key move is that we compute symbols instead of only tagging emotions.
The visual on the screen is the narrative petri dish. It places emotion, words, images and character relations into one computational field, then lets them connect, drift and recombine until they become usable stories, shots and design directions. VSA lets symbols bind as high-dimensional vectors; VSM keeps emotion, motifs and story elements in one shared space. I only need one sentence here: we are turning vague resonance into a structure that can be recalled, recombined and delivered.
This is not only a diagram. Imagine a brief from a coastal council, a museum, or a heritage attraction that wants younger families to care about a local place. The input should not be “make it nostalgic.” The system decomposes public cultural signals into motifs, character relations, emotional trajectories and visual directions.
The sample on the left shows that effect. It is not a collage of materials. It moves from safe group-level resonance toward a visible sample through an algorithmic path. For a client, the value is direct: early development no longer depends only on intuition. The system can quickly produce story assets that are discussable, editable and safe to hand into production.
The first case is an AI long-form series for a UK family audience. The world could be built from safe sources: public-domain folklore, canal and railway heritage, coastal rescue stories, museum archives and licensed local-history material. The series does not need to invent every episode from scratch. The world, relationships, places and rules first sink into a mother body, and each generation draws from the same base.
The two formulas on the left stand for two uses. The scene-generation algorithm decides whether a scene belongs on a pier, in a railway arch, beside a canal lock, or inside a small local museum, and what emotional weather it should carry. The storyboard algorithm breaks it into shot order, rhythm, character movement and visual focus. That lets a long series keep producing while staying traceable and culturally safe.
The second case is feature animation. In a British family-film setting, the character might be a child, a retired station keeper, a museum volunteer, or a small creature attached to a town landmark. The left side is not only showing pretty images. Two formulas are at work: micro-action generation and world projection. The system recalls from character, emotion and world libraries, then turns a clear story state into movement and image.
Then the algorithm makes practical decisions: should the character hesitate before entering the village hall, reach for an old ticket, look back at the pier lights, or hide a broken object in a museum drawer? What prop, colour, space and atmosphere should carry that feeling? The result is still emotionally legible, but it stays in the safer territory of character craft, family storytelling and place-based imagination.
So what is the advantage? First, we know earlier why an audience may be moved, instead of guessing after the film is made. Second, the same emotional base can grow characters, series, short films and brand content. Third, every project sends data and feedback back into the base, so the next project becomes sharper.
We are not fighting over who renders faster. We are doing the upstream work: turning audience emotion into story assets that keep producing. Others burn inspiration. We want to make inspiration into an industrial line.
The team is intentionally compact. Dai Shang leads product direction, algorithmic structure, IP method and commercialisation path. Chunling Wu leads audience research, material organisation and project work. Jianying Dou leads visual assets, storyboard direction and design delivery. At this stage, we are not trying to look large. We are trying to keep database, algorithm and delivery tightly connected.
The next stage has three priorities. First, data expansion: more public cultural material, licensed audience research, heritage archives and commercial cases. Second, algorithm calibration: make motif bonding, story generation and sample output steadier and easier to explain. Third, B-side pilots: start with safer contexts such as tourism, museums, family entertainment and place-based brands, then let the feedback flow back into the foundation.
Finally, the ask. What we need most now is not noisy expansion. It is a thicker database and a more accurate algorithmic base. The money mainly goes to three things: buying licensed audience research and industry cases, building safe cultural and narrative libraries, and continuing to tune symbolic computation and story generation. We are open to discussing a seed round, at a pre-money valuation of RMB 20 million.
We are looking for the first money willing to let this system grow in a responsible way: public or licensed data, group-level insight, clear human review, and stories that help places and characters feel alive. Thank you.