Models · 24 September 2026 · 3 min read
The models behind your video, and the four quality levels
Low, Medium, High and Max, which models sit at each level today, and how a new model joins the roster without you having to follow the news.
By The Crewnema team
The best video model changes every few months, and a business owner should not have to keep up with it. Crewnema picks the models for you, and gives you one control: a quality level. This post explains what stands behind it today. It describes the roster and how it works; it does not rank the models, because any ranking would go out of date and some of these have only just been added.
The four levels
You choose a level for each video, and High is the default. Each level pairs a video model with image models for the still pictures that a video is built from:
- Low: Gemini Omni Flash 1.1 for motion, with Ideogram V4 for stills made from words.
- Medium: Kling O3 Pro for motion, with Seedream v5 Lite for stills.
- High (the default): Seedance 2.0 for motion, with Seedream v5 Pro for stills.
- Max: Seedance 2.5 for motion, with GPT Image 2.5 for stills, using its more detailed setting for crowded frames.
The free plan is held to its own level, which uses MiniMax H3 Max for motion. It is built to let you try the whole process, from script to storyboard to a first video, at a cost that can be given away.
Why stills matter as much as motion
A video is made from pictures first: a first frame and a key frame for every shot. Where a face or a set has to stay the same, those pictures are composed by an editing model that can take your reference photos, so the model that makes motion is starting from the right person and place. That is why the level you pick sets both kinds of model.
Drafts before approval
Work you have not approved is made at a lower-cost draft quality. When you approve, the video is made at the level you chose. Every step shows what it will cost before it runs, and the wallet refuses a step it cannot cover, so a draft is a way to check the story and framing cheaply and not a surprise on a bill.
Which level should I pick?
- Testing an idea, or a video that will only be seen on a small screen for a day: a lower level is enough.
- Anything customers will see repeatedly, or that carries your brand: High.
- A hero advert where detail matters, such as a crowded scene: Max.
Treat these as starting points and look at the first draft. The level changes how much the video costs and how it is made, and your brief and photos change the result more than the level does. See how to get high-quality AI video.
What happens when a new model arrives
A new model is added by reading the provider’s own published description of it, including exactly which inputs it accepts, rather than from memory. Its request is checked against that description before anything is sent to it, and it only starts to be used for a level once that is done. You do not need to change anything: your level stays the same and the model behind it can improve. When a model joins or moves between levels in a way you would notice, we will say so here.
Make your first video
Crewnema is in private beta and lets businesses in a few at a time. Request access and we will write when yours is open.