Bring myths to life with Samsar

Myths shape our world, epics passed down through generations define our way of life even today. Some myths survived thousands of years while others faded through time..

True interpretation of myths is however a lost-art. The originals are dense and in languages obscure, which leaves plenty of room for interpretation. Back in the day, learned scholars were tasked with interpreting the mythologies and dispersing the essence to the commons. Commercial movies and tv-shows however are optimized for engagement and hardly ever do a faithful representation of the original texts.

Enter Samsar - Prompt to Visualize anything

Generative Large Language models have over millions of training runs absorbed the essence of each myth, as well as the tertiary literature, derivative works and much more.

LLMs differ in the ways they interpret myths according to their personality and training too, but mostly stay true to the source matter and unless specifically instructed otherwise, are not hyper-optimized optimized for engagement.

We will be doing a few prompt to video demoes based on the popular Chinese novel - Journey to the West. Each demo will be rendered via one of two LLM backends - GPT 5.6 Sol or Qwen 3.8 Max.
For each of these renders, we use Cosmos 3 Super for Video and Seedream for image generation stage.

Journey to the West - Four Generative Episodes

To start off, we will be creating the well-known episode where Wukong storms the Celestial Palace after rebelling against the Jade emperor. With Qwen 3.8 inference, the theme and visuals are true to the original, the speech dialogs are natural and stays true to the original texts. the animations are more dynamic and follow the storyline accurately.

Qwen 3.8 creates striking image prompts, and very dynamic meta prompts for the AI Video stage, which sometimes results in camera transitioning from the subject of current scene to a preview of the subject in the next scene. It also feels more authentic for this theme, the descriptions of the visuals and the dialogs seem a bit more authentic for the current theme.

GPT 5.6 Sol creates prompts for crisp animations, no more than needed, no extra camera transitions, it preserves the physics better and the animations seem very realistic. In this episode, where Wukong battles the Bull Demon King. The animations and physics are near-prefect without any extra movements.

For the final showcase, we will be following Sun Wukong as he goes on a journey to the Lion Camel Ridge to rescue his master.
Once again uses GPT 5.6 Sol, you can really tell by how well the storyline is structured, the animations are great, the visuals and theme adherence is accurate and to-the-point too.

These renders are more context and prompt adherent and physics and animations much better than similar demoes from even 6 months ago. The combined improvements in capabilities across modalities combined with improvements in the Samsar Framework to bring out the best of these models now really shows. Few months earlier we would frequently re-roll one or more scenes before final publish, now we hardly ever need to.. The renders are just so good in 1-shot.
There are two more llm inference models available in the arsenal which you can try creating 1-shot renders with, Kimi K3 and Gemini-3.1.
Each of these LLM backends create its own unique interpretations for the mythology with subtle differences that reflect their own personality. All renders created with the framework are a wonder and joy to behold.

Still just scratching the surface though, you could create videos from mythological stories all day long, create series and episodes or entire channels or apps just based on a single epic !!!
Always wanted to start that ancient myths channel, but weren't sure how? Just head over to the production web-app, choose your preferred model settings (Or use defaults) Ask the assistant for prompt ideas, iterate, finalize prompt and render. Then wait for the system to generate the perfect visualization for your favorite mythological stories in the model settings of your choice.
Contact at hello@samsar.one for customized or enterprise integrations.