Correction, 2026-10-02: this page assembled the trailer with Merge Videos. Merge Videos (the Montage node) returns the joined picture without sound, which would drop the audio bed the whole trailer is cut to, and it takes at most 10 clips. Assembly now goes on the Video Editor timeline, which is not a priced tool, and the 60-second total is about 768 credits rather than 778.
A trailer is a sound design problem with pictures attached. The structure, the rhythm, the rising tension and the hits are all carried by audio, and the visuals are cut to them. That makes trailers a genuinely good fit for generated video, because short disconnected shots cut to a sound design bed is exactly what generation produces well.
TL;DR
- Sound carries it. Build the audio bed first, then cut pictures to it
- Short shots. 1 to 2 seconds each, which is inside where generated video holds
- Cheapest shots: Pruna p-video Draft at 4 credits per 5 seconds (720p), cut down to the beat; Wan 3.0 at 480p is 24 if you want Wan
- A full 60-second trailer: about 770 credits, roughly $7.70
- No text in the generated frames. Title cards go in the edit
The structure that works
Trailer structure is conventional because it works, and the convention is about escalation:
- The quiet open. One image, one sound, establishing tone. 3 to 5 seconds.
- The setup. What world are we in. Three or four shots, held slightly longer.
- The turn. Something changes. A single hit, often with a beat of silence before it.
- The escalation. Cuts get shorter, sound gets denser. This is most of the runtime.
- The hard stop. Everything drops out. One image or one line, in near-silence.
- The title card, then a final short hit.
Sixty seconds total for most purposes. The escalation section is where generated footage does its work, because the cuts are short enough that nothing has time to fail visibly.
Build the audio first
This inverts how most people approach it and it is the single most useful thing in this article.
Lay down the rhythm before you have any pictures. Risers, impacts, a low pulse, and the silences. Then cut shots to the hits you already placed.
Canon 8frame prices, a credit is $0.01 at pack rate:
| Element | Route | Credits |
|---|---|---|
| Impact or hit | ElevenLabs SFX | 1 to 2 |
| Riser | ElevenLabs SFX | 2 to 3 |
| Ambient bed | ElevenLabs SFX (up to 22 s) or Lyria 2 (30 s) | 6 to 13 |
| Music bed, 1 minute | ElevenLabs Music | 81 |
| Score for a full trailer, exposed | ElevenLabs Music, 1 min | 81 |
| Voiceover line, 1,000 characters | ElevenLabs TTS | 13 |
Eight hits and four risers is 60 credits, about sixty cents, and that plus a music bed is the skeleton of the whole thing.
The shots
Short shots mean the cheap tier is the correct tier. You are cutting 5-second generations down to 1 to 2 seconds, so resolution and long-form coherence both matter less than usual.
| Element | Route | Credits |
|---|---|---|
| Shot, 5s at 480p | Wan 3.0 | 24 |
| Shot at 1080p | Wan 3.0 base, 5s | 95 |
| Hero shot with controlled motion | Kling v3 Standard silent, 5s | 57 |
| The one shot that carries it | Veo 3.1 Fast, 8s | 168 |
| Long establishing plate, 30s at 480p | Wan 3.0 base | 142 |
| Assembly, cut to the audio bed | Video Editor timeline | 0 (not a priced tool) |
A 60-second trailer: 20 shots at 480p (480), one hero on Veo 3.1 Fast (168) and audio (~120) is about 768 credits, roughly $7.70. Inside a $19 Starter month. Cut it on the timeline of 8frame's Video Editor mode, with the audio bed on sound layers. Merge Videos on the canvas does not fit a trailer: it joins at most 10 clips and returns the picture without sound.
Why short cuts save you
Generated video degrades with duration: morphing accumulates, flicker becomes visible, and the eye gets time to notice what is subtly wrong. A 1.5-second cut gives none of that time.
This is why trailers are an unusually forgiving format. The same footage that would not survive a 6-second hold reads as intentional inside a fast cut sequence.
Practical rule: generate 5 seconds, use the best 1.5. You will discard most of every clip and that is correct.
Title cards and text
Do not generate text inside the frames. Generated type garbles, changes between attempts, and for a title card that is fatal.
Generate clean plates with a visually quiet area, then set the type in your editor with your real font. If a card needs text integrated into the scene, Nano Banana Pro at 21 credits handles it as a still, and you should read it back from the image rather than trusting the prompt.
What to avoid generating
Faces held long. Keep them brief, distant, partial or absent. A face in close-up for three seconds is where realism fails.
Hands doing something specific. Reliable failure.
Crowds. Bodies intersect.
Two people interacting. Same.
Camera moving through architecture. Straight lines bend because the model has no 3D scene.
Trailers can be built entirely from shots that avoid all five: landscapes, weather, objects, light, textures, distant figures, abstract motion.
The resolution note
There is no video upscaler on this canvas, so generate at your delivery resolution. For 1.5-second cuts inside a graded and grained timeline, 480p works in a 1080p delivery. For the one hero shot that holds, generate at 1080p.
FAQ
What is the best free AI trailer maker? A free editor plus generated shots. The tool that matters is the editor, because trailers are cut, not generated.
How long should shots be? 1 to 2 seconds in the escalation. Generate 5 seconds and use the best 1.5.
How much does a trailer cost? About $7.70 in credits for a 60-second trailer with 20 shots.
Can I generate the title card? Generate the plate, set the type in your editor.
Build the sound, cut pictures to it, keep every shot short. The 8frame canvas is free and unlimited; generation is paid from $19/month.