Previz on a Budget: A Guide for Indie Filmmakers

Most indie films don't have storyboards. Not because the directors don't want them, but because professional storyboarding has always required a significant investment—one that most micro-budget productions simply can't make. A full feature storyboard can run $5K–$25K and take weeks to produce. That money and time don't exist on most indie budgets.

So the storyboard gets skipped, and the director walks onto set with a shot list and a prayer.

AI tools have changed this. Not in a "the future is here" way. In a practical, boring, useful way: you can now generate storyboard-quality frames for your film in a day, for the cost of a monthly subscription. The output isn't perfect. But it's good enough to plan your shoot, brief your DP, and stop wasting setups on coverage you'll never use. For productions that would otherwise have no previz at all, that's a meaningful shift.

The case for previz when you have no money

It seems backwards, but the tighter your budget, the more you need previz. Studios can afford to shoot extra coverage and figure it out in post. If you're burning through a $30K budget in 12 shoot days, every wasted setup costs you something you can't get back.

Previz forces decisions. Which angles do you actually need? Where does the camera go first? What can you cut from the shot list without losing the scene? These are questions you want answered at your desk, not on location with your crew standing around.

It also solves the communication problem that kills indie sets. When the DP asks "what are you thinking for this scene?" and you hand them a storyboard, you've just saved 20 minutes of verbal back-and-forth and eliminated a round of "that's not what I meant."

What traditional previz requires

Professional pre-visualization has always been an investment. Some rough numbers:

These costs are justified—good previz pays for itself many times over on set. But for a first-time filmmaker working with their own savings, or a film student stretching a grant, this level of investment simply isn't an option. The result is that the productions with the smallest margin for error are the ones that go in with the least preparation.

What's different now

AI image generation got good at cinematic composition somewhere around 2024–2025. The models understand things like lens choice, depth of field, and motivated lighting because they were trained on millions of film stills and production photographs. You describe a "medium close-up, warm tungsten practicals, 35mm film grain" and the output actually reflects that.

The meaningful change is that previz is now accessible to productions that couldn't afford it before. If your choice used to be "full storyboard or no storyboard," there's now a middle path: AI-generated frames for $20–$50/month that you can produce in 1–3 days.

The other shift is iteration. In any traditional workflow, revisions cost time and money, so you tend to accept "good enough" early. When generating a new version takes seconds, you explore more freely. You try the tighter framing. You test the lower angle. You discover things about your own visual instincts that you wouldn't have found if every iteration carried a cost.

A workflow that actually works

1. Do the scene breakdown first

Don't start generating images immediately. The AI is a rendering tool, not a thinking tool. You need to know what you're asking for.

Go through your script scene by scene. For each one, write down: the location, time of day, who's in the scene, what physically happens, and what the emotional shift is from beginning to end. Identify the 3–5 key moments per scene that would need their own frame.

This work is free and it's the most important step. A good scene breakdown fed to any AI tool will produce usable output. A lazy one will produce generic frames you can't use for anything.

2. Pick your visual references early

Before you generate a single frame, decide what your film looks like. Not in a vague "I want it to feel real" way. In a specific way.

Pull stills from films you're drawing from. If you're making something with the observational distance of Kelly Reichardt, that's a different visual language than the handheld claustrophobia of a Safdie brothers film. Write down your default lens range, your lighting approach, your color temperature. These become your generation parameters.

"A dimly lit kitchen, 35mm film grain, warm tungsten tones, medium close-up" produces something you can work with. "A kitchen scene" produces nothing.

3. Generate and iterate scene by scene

Work through your script in order. For each key beat, generate a frame. Look at it critically: is the framing right? Is the camera where you'd actually put it? Does the lighting match what you'd get on location?

If something's off, adjust the description and regenerate. Push the camera tighter. Try it from a lower angle. Change the lens from a 35mm to an 85mm. Each iteration takes seconds.

4. Put it together and share it

Export the frames as a PDF or an animatic with rough timing. Share it with your crew before the first production meeting. When everyone shows up having already seen the same movie in their head, prep meetings get shorter and shoot days get smoother.

5. Let the storyboard go on set

Your storyboard is a plan, not a promise. When you get to location and the room is smaller than you imagined, or the light is coming from the wrong side, or your actor does something unexpected and better than what you planned—adapt. The storyboard already did its job. It made you think through every scene in advance. That preparation stays with you even when the specifics change.

What doesn't work well yet

I'd rather be upfront about the limitations than have you find out mid-prep:

Multi-character scenes are unreliable. Two people in a frame work fine. Five people at a dinner table, with specific blocking, where their spatial relationships matter? The AI struggles with that. You'll need to either simplify or sketch those by hand.

Continuity across frames takes effort. Filmvision works to keep characters and settings consistent across your storyboard, but it's not perfect—especially over long sequences or when costumes and lighting shift between scenes. Review your boards with continuity in mind, and flag frames where something drifts. The tool gives you a strong starting point, but your eye is the final check.

It won't teach you how to frame a shot. The tool generates options. If you can't tell the difference between a good frame and a mediocre one, the AI won't help you learn. Spend time studying cinematography. Watch films with the sound off. The tool is only as good as the person directing it.

If your budget is next to nothing

Minimum viable previz for very little:

1. Scene breakdown by hand. Paper or a free doc.
2. Filmvision for your 10–15 hardest scenes—the ones with complex blocking, the ones at locations you can't revisit, the ones where the emotional stakes demand precision.
3. Stick figures for everything else. They worked for Spielberg on Raiders and they'll work for you.
4. Assemble into a PDF. Any tool. Google Slides, Keynote, Canva, whatever.

You don't need to previz every frame. Focus on scenes where getting it wrong costs you the most.