Storyboarding With AI Stills Before You Render

The economics that make boarding obvious

A 15-second video costs 30 credits. A still costs one, and the first three each month cost nothing. That's a thirty-to-one ratio between deciding and committing, and it's the entire argument for boarding: every decision you can move from the expensive step to the cheap one is worth moving.

The traditional reason for storyboarding — coordinating a team — doesn't apply to a solo creator. The economic reason applies more strongly than it ever did in hand-drawn animation, because the cost difference between a sketch and a rendered scene has never been this explicit.

One frame per beat

Break the piece into beats: the moments where something changes. A character arrives, a problem appears, a solution is tried, it fails, it works. One frame each.

If a beat needs two frames to be legible, that's a sign it's actually two beats. If two adjacent frames look nearly identical, that's a sign one of them isn't a beat at all — which is exactly the discovery you want to make at one credit rather than thirty.

For a 60-second piece, four to eight beats is typical. Substantially more than that usually means the script is doing too much.

Board in the ratio and style you'll publish

A board in the wrong aspect ratio teaches you the wrong things about framing — what fits, what crowds, where the subject sits. Generate boards at the video's ratio so the composition you approve is the composition that's achievable.

The same goes for style. Boarding in a different style than you'll render in makes the board prettier and less useful; the point is to preview the actual thing. Use the storyboard workflow in the same named style the video will use.

The cut is the point

Lay the frames out in order and read them as a sequence. The row will tell you things a script won't:

• Which beat is redundant, because the frame before it already said that.
• Which transition doesn't work, because the two frames have no visual relationship.
• Which beat is missing, because the jump between two frames is too large.

Then cut. A boarded sequence that loses two of seven beats before rendering has saved 60 credits and, more importantly, produced a tighter video. Cutting after rendering saves nothing and hurts more.

Boards become opening frames

Here's where boarding stops being planning and becomes production. An approved board frame can be used as the first frame of the generated scene it represents — which is the most direct control available over how a clip starts.

This matters most for longer pieces. Videos beyond 15 seconds are produced as chained scenes, where each clip's final frame seeds the next one's opening. The chain's coherence depends heavily on where it starts. Beginning from a frame you deliberately approved, rather than one the model improvised, propagates through everything after it.

Board with the chain in mind

Because scenes chain, adjacent frames should be visually continuous — same location, compatible lighting, the character in a plausible next position. A board that jumps between unrelated compositions will produce a video that feels like unrelated clips, because that's what it asked for.

This is a constraint worth designing around rather than fighting. Sequences built from small changes between adjacent frames chain far more convincingly than sequences built from dramatic cuts, which is a real stylistic consequence of how the pipeline works.

What to do when the board is wrong

Regenerate the frame, not the plan. The most common mistake at this stage is deciding the whole concept is broken because one frame came back poorly. Frames are cheap and prompts are adjustable; concepts are the expensive thing to replace.

If three attempts at a beat all fail, that's different — it usually means the beat is under-described rather than unrenderable, and rewriting it in concrete nouns fixes it more often than rerolling does.

When not to board

Short pieces don't need it. A single 5-second scene has one beat and boarding it is just generating it twice. The break-even is somewhere around three scenes — below that, the decision between still and video matters more than the boarding does.