Aspect ratio and resolution are the two settings people leave on the default and then regret. The wrong ratio means cropping away most of a shot you paid for. The wrong resolution means either a soft final video or a draft that cost five times what it needed to.
Both decisions are easy once you start from where the video will be watched.
Start from the destination
| Where it will play | Aspect ratio |
|---|---|
| TikTok, Instagram Reels, YouTube Shorts, Stories | 9:16 (vertical) |
| YouTube, websites, presentations, TV | 16:9 (horizontal) |
| Square feed posts and ads | 1:1 |
| Portrait feed posts | 4:5 |
| Cinematic and banner formats | 21:9 |
Generate at the ratio you will publish. It is tempting to make one wide video and crop it for every platform. But a 9:16 crop from a 16:9 frame keeps only about a third of the width. The model framed its subject, its camera move and its background for the wide shot, so the crop cuts through all three, and the resolution drops with it.
If you need the same idea on several platforms, generate it once per ratio. Doing that from the same source image keeps the versions consistent; see the image-to-video guide.
Which models support which shapes
Most video models on TurboMax offer 16:9 and 9:16. Beyond that they differ:
| Model | Aspect ratios |
|---|---|
| HappyHorse 1.1 | 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 21:9, 9:21 |
| Seedance 2.5 / 2.0 / 2.0 Mini | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 |
| MiniMax H3 | 21:9, 16:9, 9:16, 1:1, 4:3, 3:4 |
| Kling 4.0 / 4.0 Flash | 16:9, 9:16, 1:1, 21:9 |
| Wan 3.0 | 16:9, 9:16, 1:1, 4:3, 3:4 |
| Grok Imagine | 16:9, 9:16, 1:1, 3:2, 2:3 |
| Kling 3.0 / 3.0 Turbo / O3, Wan 2.6, PixVerse V6 | 16:9, 9:16, 1:1 |
| Veo 3.1 / 3.1 Fast | 16:9, 9:16, Auto |
| Gemini Omni 1.1 Flash | 16:9, 9:16 |
Two practical consequences:
- For 4:5 portrait feed video, HappyHorse 1.1 is the model with a native option. On anything else, generate at 9:16 and crop the top and bottom slightly, which loses far less than cropping a wide frame.
- For 21:9, use Seedance, MiniMax H3, Kling 4.0 or HappyHorse.
On image-to-video, the source image's shape matters too. Give the model a picture at the ratio you are asking for. A square photo animated at 9:16 has to be extended or cropped before the motion even starts.
Resolution: what it costs and when it matters
Most video models are priced per second, and many have a separate rate for each resolution. The difference can be large. At the time of writing, a second of Seedance 2.5 at 1080p costs more than five times a second at 480p, and on Kling 4.0 1080p costs about one and a half times the 720p rate. A few models, like Veo 3.1, charge almost the same at either resolution. The models page shows the current rate for every model and resolution.
Resolution options also vary:
- Down to 480p or lower for drafting: Seedance 2.5, Seedance 2.0, Seedance 2.0 Mini, Wan 3.0, Grok Imagine, Grok Imagine Video 1.5, Vidu Q4 (540p) and PixVerse V6 (from 360p).
- 1080p finals: Kling 4.0, Seedance 2.5 and 2.0, Wan 3.0, Veo 3.1, Kling O3, Kling 3.0 Turbo, HappyHorse 1.1, Wan 2.6, Vidu Q4, PixVerse V6, Gemini Omni and Grok Imagine.
- MiniMax H3 uses its own tiers, 768P and 2K.
When 720p is enough
Small placements in a feed, a preview loop on a web page, a reaction clip, or anything that will be shrunk into a corner of a larger edit. Most phone screens show in-feed video well below full size.
When 1080p is worth it
Full-screen vertical video that people watch on a phone, YouTube, anything shown on a laptop or TV, footage you will crop or reframe in an editor, and anything a client will judge on detail.
Draft low, finish high
The expensive part of AI video is not the final render; it is the attempts before it. Most of what you are checking in a draft, like composition, motion, timing and whether the idea works at all, is visible at low resolution.
- Fix the frame as a still in an image model, where an attempt costs a small fraction of a video attempt.
- Draft the motion at the model's lowest resolution, at the duration you need.
- Lock the prompt. Change one thing at a time until the draft is right.
- Render the final once, at 1080p, with the same prompt, ratio and source image.
Note that a higher-resolution render is a new generation, not an upscale of the draft, so it will not be frame-identical. Keeping the prompt, ratio, duration and source image the same keeps it as close as possible. Approve the idea at draft quality, and expect some differences in the final.
Duration multiplies everything
On per-second models, cost scales with length as well as resolution. A 10-second clip costs twice a 5-second one at the same settings. Two habits help:
- Generate the length you will use. If the edit needs 4 seconds, do not render 15 and trim.
- Mind each model's range. Veo 3.1 runs 4 to 8 seconds; Kling 4.0, Seedance 2.5, Wan 3.0 and Grok Imagine go up to 30. For a longer piece, it is usually better to plan several shots than to stretch one generation.
For a fuller breakdown of what drives price, see how much AI video generation costs.
A quick checklist
- Where will it play? Set the aspect ratio from that.
- Is the source image already at that ratio?
- Does the model support the ratio natively?
- Drafting? Use the lowest resolution offered.
- Final? 1080p for full-screen or large displays; 720p for small placements.
- Is the duration the length you will actually use?
Generate vertical, square and widescreen AI video on TurboMax AI with Kling, Seedance, Veo, Wan, MiniMax, HappyHorse and more. Pick any ratio and resolution, see the price before you generate, and pay from one prepaid balance. No monthly subscription.