Field notes
AI filmmaking, tested in production
Practical guides, model comparisons, and production lessons from building cohesive films with generative video.
Featured · AI writing modelsWe tested 8 LLMs on 104 screenplay scenes. Claude and Kimi led the fieldA controlled OpenRouter experiment comparing Claude Opus 5, GPT-5.6 Sol, Kimi K3, DeepSeek, Nemotron, Inkling, Gemini, and Qwen—with every generated script published.Read the analysis →Latest articles
4 articlesAI image modelsWords, references, or AI descriptions? We tested style consistency 252 timesFive controlled, blinded image-generation experiments across radically different target styles—with every result shown side by side.AI video modelsMaking AI storyboards reliable: five rounds of experiments with Seedream and SeedanceHow we tested one-canvas storyboard sheets for AI filmmaking — prompt registers, sequential mode, vertical formats, and panel detection — with every result measured.AI video modelsHow we compare AI video models for filmmakingThe rubric Sundream uses to judge video models for narrative work — consistency, direction-following, and cost per usable shot.