Skip to main content
Ask GoFrame Editor Agent for a shot and it generates the clip, saves it to the Media pool and places it on the timeline. You can start from a text description, animate a still, keep a character consistent with reference images, or restyle footage you already have.

​
Before you start

  • Credits. Video is billed per second of output and is the most expensive thing you can generate. A 5-second clip costs between 115 and 375 credits depending on the model (see Options). The agent shows the cost and waits for your approval before a paid run, unless you’ve set that skill to auto-approve in Settings → Agent → Permissions.
  • A plan that includes the model. Veo 3 Fast, the default, carries the pro badge and needs a paid plan. Models with a pro badge in the model picker are not available on Free.
  • Concurrency. Free runs 1 generation at a time, Creator 2, Pro 4. Extra requests wait in line.
  • Inputs (optional). A start image, an end image, reference images, or a source video from the Media pool. Mention them with @ in the chat.

​
Ask the agent

Type what you want or start with /video-gen. Describe the subject and its action first, then the camera, then the look. Say the duration and aspect ratio if the defaults in the generation settings aren’t right.
The generation settings (the sliders icon in the composer) set the defaults for every request: an aspect chip (Wide 16:9, Vertical 9:16, Square 1:1, Classic 4:3, Tall 3:4, Social 4:5, Cinematic 21:9), Duration (Auto, 5 or 10 s; Veo 3 Fast uses 4, 6 or 8 s) and Sound on / Sound off. Pick the model with the ✦ button in the composer or /model.

​
Options

Which one to pick:
  • Veo 3 Fast for the default balance of quality, speed and sound, and for first-and-last-frame transitions.
  • Kling O3 Pro to edit or restyle existing footage, or to keep a character consistent from reference images.
  • Kling 3.0 Pro for cinematic motion and precise camera control.
  • Seedance 2.0 for the longest clips (15 s), multi-shot narratives and rich references. Use Seedance 2.0 Fast while iterating and the standard tier for the final render.
  • Kling 2.5 Turbo and Seedance 1.0 Lite for cheap drafts.
Durations are hard limits. Ask for 6 seconds on a Kling model and the agent rounds to 5 or 10. Square (1:1) is not available on Veo 3 Fast; 21:9, 3:4 and 4:3 exist only on Seedance 2.0. If a model is temporarily unavailable on the provider’s side, the run fails and the agent offers alternatives: try another model, or wait and retry the same one later.

​
Where the result lands and what to do next

  • The clip is saved to the Media pool with the Created by AI agent badge, and its prompt, model and variant are recorded under AI-generated in the info panel.
  • By default it’s placed on a video track and reported in a Timeline changes card. Say “at the playhead”, “after the intro” or “just to the library” to change that.
  • A Video generation job card in the chat shows progress; when it finishes you get a preview with Reveal in library and Regenerate. Regenerate re-runs it with the same parameters, which is the quickest way to get a second take.
  • Starting from an image you just generated? Click the Animate it into a video clip suggestion chip under the agent’s reply.
  • Silent clip? Ask for sound from picture: see Music and sound effects. Too soft for a big export? See Upscale. Need a sequence rather than one shot? Run /director and let the agent plan, generate and assemble the whole thing.
A generated clip placed at the end of the timeline, with the Media pool info panel showing its AI-generated provenance

​
Do it from the Storyboard

The chat is the primary path, but the Storyboard tab works for visual exploration: switch the prompt bar to video with the type icons, choose an aspect from the aspect menu, describe the shot, and finished clips appear as nodes on the board that you can compare and bring to the timeline. See Storyboard.

​
Prompt tips

  • One clip, one main action. Describe the subject’s movement first, then the camera move (“slow push-in”, “orbit”, “handheld follow”).
  • Use time brackets for pacing: “[0–3 s] wide shot of the harbor, [3–6 s] cut to a close-up of the rope”. Models follow this better than “first… then…”.
  • For dialogue, give the exact line in quotes and the tone. Keep lines under 15 words; English renders most reliably.
  • Say what to avoid: “no text, no subtitles, no watermark, no extra limbs”.
  • For a recurring character, reuse the same reference image and the same full description in every prompt. Each generation starts from scratch.
  • Run /visual-identity once and every clip in the project shares the same palette and camera language.