シングルエージェント自動ワークフロー — Glif とチャットしてビデオを取得
Meta-pattern: "no-prompt" workflow. a16z partner Justine Moore (@venturetwins, 173k followers) shows how a single agent (Glif) abstracts the two-step pipeline behind a chat interface. The cost of convenience: less control + agent markup. Useful as a baseline for first-time users who want to test the pipeline without learning prompt craft. Source: EvoLinkAI/GPT-Image-2-Seedance2-Workflow Case 18.
The prompt chain
Copy these prompts, swap your character and topic.
Step 1GPT Image 2· plain· 1 line
I want a 10-second social video. Reference image attached: [your character or product]. Make it feel like a [dance reel / cooking demo / product showcase / etc]. Vertical 9:16. The agent will translate this into the GPT Image 2 storyboard prompt and the Seedance animation prompt automatically.
No manual prompt — the user describes their goal in natural language to the Glif agent, which internally generates the GPT Image 2 prompt for the storyboard.
Step 2Seedance 2.0· plain· 1 line
I want a 10-second social video. Reference image attached: [your character or product]. Make it feel like a [dance reel / cooking demo / product showcase / etc]. Vertical 9:16. The agent will translate this into the GPT Image 2 storyboard prompt and the Seedance animation prompt automatically.
Same agent that wrote the GPT Image 2 prompt also writes the Seedance prompt and runs it. End-user sees only the final video. Cost: agent platforms add a markup on top of base model costs.
Your turn
Upload your character and run this pipeline.
The prompts above are pre-filled in the workbench. Swap in your image and tweak the topic — first generation is free.
Try this technique →