My idea-to-story pipeline is getting better thanks to Opus 4.8 and GPT-5.5. I basically work with them as a co-writer/director, and then you have image and video models getting better and cheaper as well, which creates these amazing compounding effects. I was able to create this short for less than $10, and it would have been even cheaper if not for some iteration. This was largely thanks to free image generation via Codex (GPT-Image-2),
@PrunaAI's p-video and p-video-avatar models, and Gemini 3.1 TTS for the audio. Gemini is solid. It's definitely on the more expressive side, maybe even a little too expressive at times, but the final output came out alright. From there, the video models can take the audio directly and generate the final shots. The mockumentary style actually works really well with p-video-avatar when it's just one person talking directly into the camera. This story is called The Last Coder, and I literally just came up with it on a walk, then had it fleshed out in a few hours. What a time to be alive!