A trailer that does not pull tension reads as a montage. Today’s video models will gladly hand you 30 cool clips with nothing connecting them; that is the default output. This tutorial builds a 45-second trailer with a real arc: a setup that poses a question, an escalation that raises the stakes, and a button that makes the viewer want the full version. The structural decisions matter more than the prompt craft. Get the arc right and even mediocre shots feel cinematic; get the arc wrong and the best AI footage still feels random.
One thing changed in the tooling since early 2026, so plan around it: OpenAI shut down the consumer Sora app on April 26, 2026, and the Sora API is scheduled to sunset on September 24, 2026, per OpenAI’s own discontinuation notice. If you are starting a trailer today, build it on Veo 3.1 and Kling 3.0 instead. The arc below is tool-agnostic; the tool picks are the parts that age.
TL;DR
- Lock a 3-act timeline before generating a single shot: 15s setup, 25s escalation, 5s button.
- Shot budget: setup 3-4 long shots (3-5s), escalation 8-12 fast shots (1.5-3s), button 1-2 held shots (2-3s).
- AI clips are short: Veo 3.1 generates 8-second base clips; Kling generates 5-10s on free, longer on paid. You assemble the arc in an editor, not in one generation.
- Generate one act per session for world coherence. Layer sound design (drone, hits, silence) before color. Grade the whole piece as one arc.
- As of June 2026, Google AI Pro ($19.99/mo) covers Veo 3.1 in the Gemini app and Flow; Kling has a free 66-credits/day tier and Pro at $25.99/mo. Suno (~v5.5) gets you to roughly 80% on a temp score.
The tools, as of June 2026
You assemble a trailer from short clips, so what matters is clip length, audio, motion quality, and cost per shot. Here is the current landscape.
| Tool | Base clip length | Resolution | Native audio | Consumer access (June 2026) | Best for in a trailer |
|---|---|---|---|---|---|
| Veo 3.1 (Google) | 8s | 720p / 1080p / 4K | Yes | Gemini app + Flow; Google AI Pro $19.99/mo, ~1,000 Flow credits/mo | Narrative shots, establishing frames, prompt adherence |
| Kling 3.0 | 5-10s (up to 2 min paid, 3 min with Extend) | up to 1080p | Yes (multi-shot sync) | Free 66 credits/day; Pro $25.99/mo (~3,000 credits) | High-motion, stylized worlds, cinematic lighting |
| Sora 2 | up to 25s (Pro API) | up to 1080p | Yes | Consumer app discontinued Apr 26, 2026; API only, sunsets Sept 24, 2026 | Legacy migration only — not a fresh start |
Practical read: Veo 3.1 is the safest default for narrative and establishing shots (strong prompt adherence, native audio, 4K). Kling 3.0 is the value pick for high-motion or stylized cuts, and its multi-shot mode keeps audio synced across a sequence. Generate by act in one tool when you can; mixing tools per shot is fine as long as your color pass unifies them at the end.
Who this is for
Indie filmmakers proving a concept before pitching, founders making a launch teaser, content creators building IP arcs across short formats, and ad teams making a 45-second hero piece for a campaign.
Before you start
- Write the one-sentence question the trailer poses. If the viewer cannot say “I want to know what happens”, the structure is broken.
- Decide tone first. Thriller, hopeful, comedic, mysterious all use different motion grammar and different cut pacing.
- Pick the visual world: time period, palette, location family. Trailers that change worlds across cuts feel like reels, not trailers.
- Decide the button. The last shot has to imply more: a door opening, a face turning, a line of dialogue cut off. Without a button, the trailer ends; it does not finish.
Step by step
- Lay the three acts on the timeline before generating anything. Block 0-15s setup, 15-40s escalation, 40-45s button, and label them. The structure has to exist before the shots, or it never will.
- Budget shots per act. Setup: 3-4 longer shots (3-5s each, slow motion). Escalation: 8-12 faster shots (1.5-3s each, rising motion). Button: 1-2 shots (2-3s each, suspended motion). Because Veo 3.1 caps a base clip at 8 seconds and Kling free at 5-10s, every act is several short generations stitched in your editor, not one render.
- Write prompts for motion energy, not just composition. Setup: “slow dolly, long lens, sustained motion.” Escalation: “handheld energy, faster moves, dynamic cuts.” Button: “stillness, locked-off frame, one moving element.” In Kling 3.0, use its multi-shot storyboard mode for an escalation run so motion and audio carry across the cuts.
- Generate per act, not per shot. Doing all setup shots in one session keeps the visual world coherent; jumping between acts mid-generation invites style drift. With Google AI Pro’s ~1,000 Flow credits roughly covering about ten Veo 3.1 Quality renders a month, budget your credits per act so you do not run dry mid-escalation.
- Sound carries half the tension. Run a single drone or pulse layer through the setup, percussive hits during the escalation, and a hard silence plus breath into the button. Native audio from Veo or Kling is a starting layer, not the finished mix. For a temp score, Suno (~v5.5) gets you to roughly 80%; license or commission the final track.
- Color-grade the trailer as one piece, not per clip. Setup slightly desaturated and cool, escalation richer and warmer, button stripped back. Color is part of the arc, and a single grade across the timeline is what hides the seams between Veo and Kling shots.
First-run exercise
- Pick an IP you know well (a book, a podcast concept, a personal project). Write the one-sentence question and the button shot.
- Storyboard 8 shots minimum across the three acts. Sketch on paper; do not start generating yet.
- Generate the setup act first (3 shots, one tool, one session). If the world does not feel like one place across all three, regenerate before moving on.
- Add sound design before color. Bare timing edit, then drone plus hits plus silence, then grade. Each layer reveals what the previous one was missing.
Quality check
- The viewer can name the question after one watch. If they cannot, the setup is too vague or too cryptic.
- The escalation actually escalates: cut pace shortens, motion energy rises, sound thickens. If the middle plays at the same energy as the setup, the arc is flat.
- The button works as a stop, not just an end: stillness, a turn, a held breath, anything that signals there is more.
- The visual world is one place, one time. Cross-cuts to other worlds must be obviously intentional (flashback grading, different palette).
- Sound design is layered, not just music: drone, foley hits, breath, silence. Music alone is a music video.
How to reuse this workflow
- Save the 3-act timeline template (15/25/5 with empty placeholders). New projects drop shots into the same skeleton.
- Build a sound-design library: 3-5 drones, 5-10 percussive hits, 2-3 breaths, one strong silence cue. Reusable across trailers.
- Keep a per-tone color preset: thriller LUT, hopeful LUT, comedic LUT. Apply once across the timeline, never per clip.
- Re-test the model lineup every 4-6 weeks. These tools ship fast: Veo and Kling both pushed major versions in the last year, and the best model for a given shot moves around. Picking per shot is where trailer quality benefits most.
Common mistakes
- Generating all the cool shots first, then trying to find a structure. The structure has to exist before the shots.
- Holding one cut pace across the whole trailer. Escalation requires accelerating cuts; sustained pace is a montage.
- Letting music do the work alone. Music is the floor; sound design and silence are the ceiling.
- No button. Trailers that end on a cool shot feel finished; trailers that end on a button feel ongoing.
- Mixing visual worlds without intent. Random shifts break trust; deliberate shifts (grading, palette, time) build it.
- Skipping the per-act generation discipline. Generating shot 1, then shot 7, then shot 3 invites style drift between cuts that have to feel connected.
- Forgetting clip-length limits during storyboarding. An 8-second Veo cap means a planned 12-second hero shot is two generations and a seam you have to hide.
FAQ
- Veo 3.1 or Kling 3.0 — which do I start with?: Veo 3.1 for narrative and establishing shots (best prompt adherence, native audio, 4K), Kling 3.0 for high-motion or stylized cuts and for its multi-shot storyboard mode. Generate by act in one tool, then unify with a single grade.
- Can I still use Sora?: Not as a consumer. OpenAI discontinued the Sora web and app on April 26, 2026, and the Sora API is scheduled to sunset on September 24, 2026. Start new trailers on Veo or Kling.
- What does it cost to make one trailer?: On Google AI Pro ($19.99/mo as of June 2026), ~1,000 Flow credits cover roughly ten Veo 3.1 Quality renders, enough for a serious 45-second piece if you are economical. Kling’s free tier (66 credits/day) handles a few short shots; Pro is $25.99/mo for ~3,000 credits.
- How long does it take?: Setup ~30 min, escalation ~60-90 min (most shots and regenerations), button ~30 min, plus ~60 min for sound and color. Budget a half day for a serious 45-second trailer.
- Music or original score?: Suno (~v5.5) gets you to roughly 80% on a temp score for cheap. For a published trailer, license a track or hire a composer; the difference reads.
- Aspect ratio?: 16:9 for film festivals and YouTube, 9:16 for social teasers. Veo 3.1 supports native vertical, so generate the framing you need rather than cropping a wide shot when the action is centered.