ChatCut generates video, images, voiceover, and music from the AI panel. This page covers the modes you can pick and the limits that apply while generating.
Which AI generation modes does ChatCut have?
ChatCut’s AI panel has two generation modes, picked from the mode switcher dropdown: Agent and Video Gen. Agent is the default mode you land in — it’s the general-purpose chat, whose generation skills cover motion graphics, video generation, and image generation. Video Gen is a dedicated mode for generating video clips with Seedance 2.0, and it’s gated to the paid Pro plan: on the Free Plan, the Video Gen send button shows a “Pro” badge, and submitting opens an upgrade prompt (“Seedance 2.0 is a Pro feature. Upgrade to Pro to generate videos with Seedance 2.0.”) instead of generating.
See also: What is the AI panel in ChatCut, How many credits does AI video generation cost, How many reference files can I use per generation.
How many reference files can I use per generation?
You can attach up to 9 reference files total to a single AI generation, with at most 3 of them being videos and at most 3 being audio files (the rest of the 9-item cap can be reference images). Reference videos and audio clips each have to be 2-15 seconds long individually, and the combined duration of all reference videos in one generation can’t exceed 15 seconds — the same 15-second combined cap applies to reference audio, checked separately from video.
Can I use images of real people as video generation references?
Yes — reference images containing real people work in Video Gen mode. If the model initially rejects a real-person image, ChatCut automatically resubmits the generation through a supported path, so you normally just see “Generating video…” continue as usual. This applies to first-frame, last-frame, and reference images.
Reference videos containing real people are different: they are not supported and the generation fails with “Seedance 2.0 does not support videos containing real people (model limitation). No credits charged. Try a different video.” Swap the video reference for still images of the person instead.
In the rare case a real-person image still can’t go through, the failed card shows “Seedance 2.0 does not support images containing real people (model limitation). No credits charged. Try a different image.” — no credits are charged either way.
See also: How many reference files can I use per generation
Why is my generation taking so long?
If it’s taking unusually long, the two causes ChatCut reports are a timeout on a complex prompt or reference set, or a temporary overload on the generation service — in both cases the attempt ends with an error message and no credits are charged. On a timeout you’ll see “Seedance 2.0 timed out (model limitation). No credits charged. Please try again.”; when the service is overloaded from high demand you’ll see a message stating it’s currently overloaded, that no credits were charged, and to retry later. Video generation itself is billed at a flat 0.6 credits per second of output, and a timed-out or overloaded attempt is never billed. If a timeout happens, retrying with a shorter prompt or fewer/smaller reference files lowers the chance of it happening again.