AI Video Studio
Cinematic AI video — powered by WAN 3.0 Prime
Which engine should you pick?
The model picker above offers 7 engines and they are not interchangeable. One renders in ninety seconds, another takes ten minutes. One costs 4 credits for a five-second clip, another 30. These are the numbers from our own production queue, not a vendor’s claim.
Updated 11 September 2026 · 6,075 production renders in the last 60 days · 9 engines measured
- Cheapest draftMiniMax H34 credits for 5s at 720p
- Fastest turnaroundWan 3.0 Prime96s median, submit to finished clip
- Highest quality scoreSeedance 2.5 Turbo81 / 100 on the blind index
- Longest single takeSeedance 2.5up to 30 seconds in one generation
Quality
v0.1 · blindWowMade Video Quality Index · v0.1 · 0 to 100 · Higher is better
Speed
Live dataMedian seconds from submit to finished clip · last 60 days · Lower is better
Cost per clip
Price listCredits for a 5-second 720p clip · Lower is better
Every engine, side by side
| Engine | Vendor | Quality | Median s | Credits / 5s | Max | Audio | In this studio |
|---|---|---|---|---|---|---|---|
| MiniMax H3 | MiniMax | 76 | 126 | 4 | 15s | Native | Use this engine → |
| Wan 3.0 Prime | Alibaba | 79 | 96 | 10 | 30s | Native | Use this engine → |
| Seedance 2.0 Mini | ByteDance | 79 | 254 | 10 | 15s | Native | Use this engine → |
| Seedance 2.0 Fast | ByteDance | 79 | 213 | 15 | 15s | Native | Not in this studio |
| Seedance 2.5 TurboPRELIMINARY | ByteDance | 81 | 627 | 16 | 30s | Native | Use this engine → |
| Wan 2.7 | Alibaba | — | 114 | 17 | 15s | Native | Use this engine → |
| HappyHorse 1.1 | HappyHorse | 79 | 110 | 19 | 15s | Native | Not in this studio |
| Seedance 2.0 | ByteDance | 79 | 260 | 20 | 15s | Native | Use this engine → |
| Seedance 2.5 | ByteDance | 80 | 386 | 30 | 30s | Native | Use this engine → |
Median is the middle time from submit to finished clip across the window, so half of all renders land faster. Credits are for a five-second 720p clip; longer clips and higher resolutions cost more. Engines marked preliminary have fewer than fifty completed renders and their numbers will move. Full prices and the rate every figure is derived from are on pricing details.
How the AI video studio works
The studio above is the generator itself: a prompt field, an engine picker, and the duration, resolution and aspect ratio the clip will ship in. Everything on screen is a consequence of those choices — the video is synthesised from the description, not retrieved from a library or edited from footage. Write the shot the way you would brief a camera operator, choose an engine, and the clip renders in a queue on the page.
Three routes in. Text to video: the prompt is the only input, on the four Seedance engines. Image to video: a start frame — your photo or a generated still — becomes the first frame and the model animates from it, on WAN 3.0 Prime, WAN 2.7, MiniMax H3 and Seedance. Reference to video: up to twelve references covering a subject, a style and a sound, so a character or product stays consistent across shots, on WAN 3.0 Prime. Each route has its own guide: text to video, image to video and reference to video.
One generation, start to finish
Describe one shot
Subject, what it does, where the camera is and how it moves, the light, the mood — in that order. One action per clip; a second action is a second clip.
Pick the engine and the format
Draft on a cheap engine at 480p to choose a direction, then re-run the winner on Seedance 2.5 or WAN 3.0 Prime at the ratio and resolution the piece ships in.
Generate, then finish
Extend a clip that needs more seconds, raise a soft one in the video upscaler, score it with a generated track, and join shots in the video merger.
Cost, labelling and what is not allowed
A 5-second clip starts at 5 credits and the credit price of your exact settings is shown on the Generate button before you commit; per-second rates for every engine are on AI video models compared. Plans renew monthly until cancelled and can be cancelled online at any time; the full terms are on pricing details.
Every clip generated here carries a signed C2PA provenance mark identifying it as AI-generated, in line with Article 50 of the EU AI Act; the mark travels with the file and anyone can read it back on the verification page. Prompts and uploads are screened before generation, a real person’s likeness needs that person’s consent, and the Content & Safety Policy sets out what may not be created here.
Frequently asked questions
What can I make in the AI video studio?
Short cinematic clips from a written prompt, from a still image, or from a set of reference images, clips and sounds. Seven engines are available — WAN 3.0 Prime, Seedance 2.5, Seedance 2.5 Turbo, Seedance 2.0, Seedance 2.0 Mini, WAN 2.7 and MiniMax H3 — generating 5 to 30 seconds at 480p up to 4K, in six aspect ratios, most of them with native audio. A finished clip can be extended, upscaled and scored without leaving the site.
Which engines generate video from text alone?
Seedance 2.5, Seedance 2.5 Turbo, Seedance 2.0 and Seedance 2.0 Mini take a prompt with nothing attached. WAN 3.0 Prime, WAN 2.7 and MiniMax H3 are built around an input: a start frame, or on WAN 3.0 Prime and MiniMax H3 a set of references. The picker shows what each engine needs before you generate.
How do I generate a video from an image?
Add a start frame — a photo you upload or an image generated on the site — write what should happen over the next seconds, choose an engine that accepts a frame, and generate. The first frame of the result is your image, unchanged. Seedance 2.5 and Seedance 2.0 also accept an end frame, so a move can be made to arrive at a second composition you choose.
What does a generation cost?
It depends on the engine, the resolution and the duration, and the exact credit price for your current settings is printed on the Generate button before you spend anything — a 5-second clip starts at 5 credits. Drafting at 480p on a cheaper engine and re-running only the winner at the resolution you ship is the least expensive way to work.
Can I make a clip longer than 30 seconds?
Yes. Thirty seconds is the longest single generation, but Extend continues a finished clip from its last frame, adding up to 15 new seconds per pass, and passes can be chained. The look carries over because the model continues the picture rather than reinterpreting the prompt. Six of the seven engines have an Extend variant; WAN 3.0 Prime clips are continued on Seedance.
Does the video come with sound?
Most engines generate native audio alongside the picture — ambience, movement, effects — and it can be switched off if you would rather score the clip yourself. WAN 2.7 uses an audio guide instead. For a composed soundtrack, generate a track in the AI music generator and lay it over the finished cut.
Which formats and resolutions are available?
Six aspect ratios — 16:9, 9:16, 4:3, 3:4, 1:1 and 21:9 — chosen before generation rather than cropped afterwards, so a vertical clip is composed vertically. Resolution runs from 480p to 4K depending on engine, and a clip that is right but soft can be raised to 1080p in the video upscaler as a separate pass.
Is the output labelled as AI-generated, and what is not allowed?
Every clip carries a signed C2PA provenance mark identifying it as AI-generated, in line with Article 50 of the EU AI Act, and anyone can read it back on the verification page. Prompts and uploads are screened before generation; a real person's likeness needs that person's consent, and the Content and Safety Policy sets out what may not be created here.
Guides and tools
One credit balance across video, image, music and the editing tools.
AI Video Generator
The full picture: engines, formats, costs and examples.
OpenText to Video
How to write a prompt that produces the shot you meant.
OpenImage to Video
Start frames, end frames and references, explained.
OpenAI Video Extender
Continue a clip past its last frame.
OpenAI Product Video
One product photo, moved on camera.
OpenVideo Upscaler
Raise a finished clip to a crisp 1080p.
OpenAI Music Generator
Score the cut with a track made the same way.
OpenVerify AI content
Read the C2PA provenance mark on any file.
Open