Compare video models and pick the best one for your goal

1Where the compare button is

In the video composer, in the row of icons next to the dictation microphone and the prompt-improvement button, there is a Compare button: on a computer it is an icon with the label “Compare”, on a phone only the icon. Hover over it and the tooltip says “Compare on 2–3 models”. Until you describe the scene it is disabled and the tooltip reminds you: “Describe the scene first”.

2How to compare models

1. Describe the scene and set the options

   Write a prompt, choose duration, quality, sound and, if needed, attach a frame or references. These settings apply equally to all models in the comparison, and the server calculates each one's price from them.

2. Press the compare button

   The Compare models window opens with the explanation: “The same prompt and settings on 2–3 models. The clips appear side by side in your feed; each model is paid separately.”

3. Choose 2–3 models

   A short list of options: the composer's current model, the fastest and cheapest ones and the strongest by quality (up to seven rows). A row shows the name, price in tokens and usual time, for example “Seedance v1.5 Pro · 10 tokens · ~50 s”. Clicking selects or clears a row; a fourth cannot be selected. Grey rows labelled “doesn’t fit these settings” do not accept your current parameters (for example the duration or a frame).

4. Check the total

   At the bottom is Total for the comparison — the sum of the chosen models. If it is above the cap, a red hint appears: “A comparison can’t cost more than 30 tokens — pick cheaper or fewer models”.

5. Press “Compare”

   The button is called “Compare · 2 models” (or 3) and shows the same sum. The window closes and the models are launched one by one.

Two models are preselected. If the first two do not fit the cap and you have not clicked anything yet, the window picks the two cheapest suitable ones: the button works right away.

3Price and the 30-token cap

Each model's price comes from the server (a price request for your settings) — it is the same sum an ordinary generation would hold. A time such as “~50 s” or “~3 min” is the median “launch → ready” in our clips over 30 days; if a model has fewer than three finished clips, there is no time.

What happens after the run: Studio puts each model into the composer one by one with your prompt and presses the ordinary “Generate”. The next model goes out only when the server has accepted the previous one. Each launch is a separate generation with its own hold and its own card in the feed, so the clips stand side by side. If sending stops somewhere (not enough balance, a limit was hit), the remaining models are not launched or paid for, and Studio says “Sent N of M — the rest didn’t run and wasn’t charged”.

4Picking a model by goal

Open the model picker in the video composer and you will see four chips above the list. Click one — the list shrinks to the suitable models, and the first gets a Recommended tag. Clicking again clears the chip. While you type in “Search models…” the chips are hidden.

Each row shows the price per second of video (“⚡ 0.9/s”) and the usual time (for example “~50 s”). The time tooltip: “Usually ready in this time (our clips over 30 days)”. Service models — lip-sync, clip editing, finishing to 1080p — are not offered in the goals. In the screenshot Grok Imagine Video 1.5 still shows the old duration limits; now it is 6, 10 or 15 s.

Related