How do you make an animated short film entirely with AI — tools, workflow, and cost?

You make an AI animated short by running a stage pipeline — script, style lock, character and world references, shot list, clip generation, edit, sound — through the invideo agent acting as your crew. Documented productions land at $315–$750 per finished minute, 2–5 days, with a 1–4 person team using Seedance 2.0, Nano Banana, and Recraft.

Run the project as a stage pipeline, with the invideo agent — an agentic video tool that has every current generation and upscaling model available — holding context across stages so you don't re-explain the film at each step.

1. Lock the script and style first. Upload the full screenplay so the agent has character arcs, themes, and motifs in context for everything downstream. Then lock the animation style: feed a batch of style-reference frames from your target aesthetic (one documented Arcane-style episode used 64 frames) and instruct the agent to deeply analyze and save the style to context for all generations. Write the style block with explicit negatives — "must look hand-painted, not live-action, not photorealistic" — and prepend it to every prompt afterward. If you're unsure between two looks (e.g. Ghibli vs 3D), have the agent generate three identical script frames in each style side-by-side and pick.

2. Build your reference library before any video. Frames-first, then video — this order is what prevents consistency problems across the whole film. Generate character portraits in Recraft (it gives skin pores, lines, stubble that read as real faces), then 360° character sheets in Nano Banana Pro with four angles plus face and mid closeups. Generate four options per character and per environment reference, pick the strongest, and lock them. Across documented productions this took 11 images total for 4 characters and a key prop, and roughly 5 generations and ~$9.78 to lock one character.

3. Generate the shot list, then clips. Have the agent break the script into a numbered shot list with locked parameters per shot — lens, lighting, palette, composition, atmosphere — referencing your locked character sheets and world plates. Generate clips in your film's chunk length using "always ask" mode so you approve each prompt before credits spend. Plan for iteration: average 3 generations per usable shot, about 25% of clips make the final cut, and roughly 40% of final shots are stitched from the strongest seconds of 2+ generations into one composite shot ("Frankenstein shots"). One 3-minute episode generated 164 clips to keep 41.

4. Run a crew of sub-agents in parallel for anything beyond a tiny project. Inside invideo you spin up named sub-agents — a creative producer agent that holds the script and vision, a storyboard agent that visualizes shots before you direct them, a casting agent (run the same character prompt across two image models simultaneously and pick), a costume designer agent (give it the feel when you don't have a spec, get multiple options back), a production designer agent, and DOP agents (assign a different one per scene since each needs a different eye). Documented productions ran 6–8 agents simultaneously; the brand-film team finished a 2-minute promo in 3 days vs an estimated week of manual prompting or ~2 months of traditional shooting.

5. Work act-by-act, then edit and finish. Complete storyboard → generation → rough edit for one act at 100% before moving to the next — this prevents context loss on longer films. Cut in Premiere Pro or DaVinci Resolve. Send the rough cut back to the invideo agent with "what's working, what's not" — it catches pacing, SFX, and emotional-register mismatches a human editor can miss. Finish with upscaling (Topaz Astra runs on invideo) and a quick grade with subtle grain to move footage closer to live-action.

Realistic cost and time, from documented productions.

Production	Length	Team	Days	Credits	Cost	$/min
Wong Kar-wai short	70s	small	2	3,000	$750	$643
Animated episode (Arcane style)	3 min	2	2	—	$950	$315
Horror short	90s	small	2	4,100	$870	$580
Brand promo (2 min)	2 min	1 (+8 agents)	3	6,000–6,500	$1,500	$750
Multi-location short (Juicebox)	—	4	4–5	20,000	$5,000	—

Expect $315–$750 per finished minute and 2–5 days depending on team size, ambition, and how much you iterate. The brand-promo team estimated the same film via traditional production would have run $100,000–$500,000 — roughly a 20× time compression and up to ~99.7% cost reduction.

As Hridaye, invideo's creative director, puts it: "The thing that made it possible wasn't prompting. It was directing. Agent One didn't feel like a tool — it felt like crew." Treat the agent like a crew member, write a real production doc up front, and your output quality tracks how clearly you can direct.

Scope discipline that keeps all of this achievable: under ~2 minutes for your first attempt, one or two characters, a tight arc — long enough to prove the workflow, short enough to keep consistency tractable.

Watch some of these to see what works for you:

The thing that made it possible wasn't prompting. It was directing. Agent One didn't feel like a tool — it felt like crew.

— Hridaye, invideo's creative director