Model guides
How to read the model pages, and which model to try first for each kind of shot.
Each page in this section covers one video model: what it can do, what creators say it is good and bad at, how to prompt it, and how to spend fewer credits on it. If you only have time for one page, read the one for the model you already pay for.
How to read these pages
Every model page has the same sections, in the same order.
| Section | What it tells you | Where the facts come from |
|---|---|---|
| At a glance | Maker, versions, length, resolution, audio, references, where to use it, cheapest way in | Official docs and pricing pages |
| Best for, Weak at | What real users report after testing | Creator posts and write ups, labelled as community |
| How to prompt it | The model's own syntax and structure | The maker's prompt guide |
| Example prompts | Prompts you can copy and adapt | Written for this site, following the official syntax |
| Settings that save credits | Cheaper ways to draft and finish | Official prices plus tested community habits |
| Known issues | Failures you will meet, and the fix | Official FAQ and community reports |
| Sources | Links to check for yourself | Official URLs |
How to read the labels:
- Official means the maker says it on its own site or docs.
- Community means creators report it. "Organic" means no sign of payment. "Sponsored" or "campaign" means the post was paid for or copied across many accounts, so treat it as advertising.
- Not confirmed means we could not find it on an official page by 15 Oct 2026.
Prices change often. Every price on these pages was checked on 15 Oct 2026. Check the pricing page before you pay.
Which model to try first
Prices checked on 15 Oct 2026.
| Job | Try first | Why | Backup |
|---|---|---|---|
| Learning on a zero budget | Veo 3.1 Lite or Fast in Google Flow | 50 free Flow credits per day, Lite costs 10 credits per clip | Grok Imagine if you already pay for X |
| Same character or product across many shots | Seedance 2.5 | Up to 50 reference files, creators rate it highest for complex scenes | Kling 3.0 with Elements |
| Action, fights, fast camera moves | Kling 3.0 | Creators pick it for action and tracking shots | Seedance 2.5 |
| Dialogue with lip sync | Kling 3.0 or Veo 3.1 | Both have native audio and a clear dialogue syntax | MiniMax H3 Max with your own voice track |
| Dialogue in Vietnamese | Seedance 2.5 | Vietnamese is on its official list of 11 dialogue languages | Record the voice yourself and lip sync later |
| Long single takes, cheap and fast | Wan3.0 | Up to 30 s per clip, 0.10 USD/second at 720p on the API | Seedance 2.5 (also 30 s, slower and pricier) |
| Atmosphere, establishing shots, B roll | Veo 3.1 | Strong ambience audio and polish | Gemini Omni Flash |
| Editing a clip you already have | Gemini Omni Flash or MiniMax H3 Max | Omni does conversational edits, H3 Max ranks first for video editing on Design Arena | Runway Aleph 2.0 |
| Running a model on your own GPU | MiniMax H3 | Open weights, runs on a 4090 or 5090 (community reports) | LTX-2.5 |
| Short loops and simple animation | Grok Imagine 1.5 Lite | From 0.02 USD/second on the API | Veo 3.1 Lite |
| HDR or color grading pipelines | Luma Ray3.2 | Native HDR with 16 bit EXR | LTX-2.5 |
Pick a model per shot, not per film
Many creators mix models in one video: Kling for action, Veo or Omni for atmosphere, Seedance for reference heavy or music driven shots. Your edit hides the switch.
All model pages
- Seedance: ByteDance, in Dreamina and CapCut. Seedance 2.5, 2.0, Fast and Mini.
- Kling: Kuaishou. Kling 3.0, and the 4.0 rollout.
- Google Veo and Gemini Omni: Veo 3.1 Lite, Fast and Standard, plus Gemini Omni Flash, in Flow and the Gemini API.
- Wan: Alibaba. Wan3.0 on the API.
- MiniMax H3 and Hailuo: MiniMax H3, H3 Max and the Hailuo web app.
- Vidu: Shengshu. Vidu Q4 Preview.
- Runway: Gen-4.5, Aleph 2.0 and the Runway plans.
- Other models: Luma Ray3.2, Grok Imagine 1.5, LTX-2.5, PixVerse, Midjourney video, and the Sora 2 shutdown.
For a side by side table with prices per clip, see Models compared.