How do you make an AI short film without writing any prompts yourself?

You make an AI short film without writing prompts by directing an agentic tool in plain language: speak or upload your story once, approve the character sheets and storyboard it generates, then let it write every generation prompt and route each shot to the right video model. One documented 3-minute short film was completed this way with zero manually written prompts.

Start by choosing an agentic workflow rather than a model interface — the invideo agent is an agentic video creation tool with all the current video models built in, and it writes its own long, detailed generation prompts from your plain-language direction. From there the workflow is five steps.

1. Load your story without writing anything structured. Use the invideo agent's audio input to verbally stream your idea — style, characters, tone — and it transcribes and organizes that into a production plan. If you already have a script, upload it whole: from an uploaded script, the invideo agent broke one short film into 20 shots across four scenes on its own, and it can ingest a brief PDF or even reference videos the same way.

2. Approve references instead of describing them. The invideo agent generates character sheets and location references from the loaded context; your job is selection. In one documented production the creator generated four options per asset and locked the best before any video generation — that approval step replaces every descriptive prompt you would otherwise write.

3. Approve a storyboard before any video credits are spent. A storyboard frame shows composition, characters, and flow for the cost of an image, so have the invideo agent storyboard the film and only greenlight panels you like. One creator's 3-minute film ran as 28 frames that the invideo agent managed end-to-end.

4. Generate with approval gates on. Switch on the invideo agent's ask-before-generating setting so it shows you each written prompt for sign-off before spending credits — you read and approve, you never author. Behind that gate it routes each shot to the right model: Seedance 2.0 for reference-driven character continuity, Kling for native multi-shot sequences, Veo where it fits — all available inside invideo, so you never pick a platform per model. In one session the agent completed 21 of 23 production tasks autonomously.

5. Fix shots conversationally. When a clip is wrong, describe the problem the way you would to a crew member — "hold on him until he lunges, no cutting" — and mark rejects with a reason; the invideo agent diagnoses the issue, proposes an alternative interpretation, and regenerates. It also flags its own physics, lighting, and geometry errors before you notice them.

The documented outcomes: the zero-prompt 3-minute film took roughly 6 hours, most of it passive generation wait time. Across five documented invideo productions, complete AI short films ran $750–$5,000 and 2–5 days depending on team and ambition — about $315–$750 per finished minute.

Watch some of these to see what works for you:

You're not writing prompts anymore. You're basically just directing the whole film.

— a filmmaker documenting an agent-based AI short film workflow