Skip to content

Create a video in one go

The Create operations take a pile of footage, stills and Kitlig Assets and hand back a finished video: an AI reads what each clip shows, writes a plan, and Kitlig builds a timeline from it and renders it to MP4. The timeline stays in the project, so you can open it in Timeline and change anything.

Find them in Sandbox under Video ▸ Create, or right-click a Kitlig Asset in Asset Mode and pick one under Create video, which opens the form with that asset already chosen.

Operation What you get
Narrated ▸ Documentary A narrated short with an arc, read by one to four ElevenLabs voices, with subtitles.
Narrated ▸ Commercial A voiced ad: hook, benefits, a close and an end card.
Narrated ▸ UGC A creator-style vertical video (testimonial, unboxing, review) in a casual first-person voice.
Movie A story cut that keeps the clips' own sound and dialogue, with a title card.
Commercial A music-driven ad with short on-screen lines instead of a voice, ending on your call to action.

The form

  • Media and Kitlig Assets: what to build from. Every image and video in a picked Kitlig Asset is offered to the AI. Pick several Kitlig Assets with Shift-click in the picker.
  • Style: the sub-category (Nature, True Crime, Product, Testimonial, Drama…) that steers the writing.
  • Idea: a sentence or two about the story or product. Leave it blank and the AI finds a story in the footage.
  • Name and Subtitle: the title card. They also name the new sequence. Leave Name blank and the AI names it.
  • Length and Frame: the target running time and 16:9, 9:16, 1:1 or 4:5.
  • Talkers, Voice style and Talker voices (Narrated only): how many voices, and which. Leave a talker's voice on "From the voice style" to use the style's voice for that talker.
  • Tone sliders: Serious to Comedic, Nice to Mean, Grounded to Stylized.
  • Subtitles: on or off, plus font, placement, size, colour and a box behind them. The title card and on-screen lines use the same font and colour.
  • Fade in and Fade out from and to black.
  • Music: pick a music file, or leave it empty and write a Music prompt to have one composed for the length. Leave both empty for no music. Music level sets how far it sits under the voice.
  • Render to MP4: turn it off to get only the timeline.

The Commercial form also has a Hero Kitlig Asset (the product the ad is about) and a Call to action for the end card.

What happens when you run it

The job shows each step as it goes:

  1. Any picked image or video without an AI description is described first. This is the same as Describe & Tag with AI, one model call per file, and the description stays on the asset. Videos over 10 seconds are split at their scene changes first, and the Settings choice to caption pictures on your computer applies here too (see Sandbox).
  2. The AI writes the plan: the lines each talker says, which shots play under them, on-screen text and the end card.
  3. ElevenLabs reads each line.
  4. Music is composed, if you asked for it.
  5. A new sequence is built, then rendered.

The sequence has these lanes, from top to bottom:

  • Subtitles
  • Titles: the title card, on-screen lines and end card
  • Pictures
  • Dialogue: Movie only, the clips' own sound
  • Narration, or Talker 1, Talker 2…
  • Music

The MP4 appears on the job's row. The plan the AI wrote is saved as a JSON file next to it, for reference.

What it costs

One planning call to your Video Assistant model (Settings ▸ Assistant), one ElevenLabs call per spoken line, one music call when you give a music prompt, and one model call for each file that wasn't described yet (text only, and much cheaper, when pictures are captioned on your computer). An AI agent connected over MCP can describe the media and write the plan with its own model instead, so only the voices and music are paid for: see Using your own AI through MCP.

Narration voices

Each voice style (Documentary Narrator, Nature (British), Indian English, Commercial (Energetic), Podcast Duo and so on) points at ElevenLabs voice IDs, talker 1 first. Set your own in Settings ▸ Narration Voices: pick from your ElevenLabs voices (press Sync first) or paste a voice ID, and add a voice per talker. Reset goes back to the voices that ship with Kitlig. A style with no voices reads with ElevenLabs' default voice, and the job says so.

Extract Selects

Image ▸ Organize ▸ Extract Selects picks the best pictures from your media and Kitlig Assets:

  • Every image is a candidate. So are frames sampled from each video, taken at the described moments first, then spread evenly across the clip.
  • Blurry candidates drop out.
  • A vision model chooses the best from labelled contact sheets, guided by What makes a select (for example "eyes open, hero product angles").

Chosen video frames are saved full-size in images/selects/. Choose a Kitlig Asset under Add picks to Kitlig Asset and every pick also joins its image pool. A short report lists why each one was chosen.

From Asset Mode, right-click a Kitlig Asset and pick Extract Selects… to run it on that asset. The picks go back into the same asset.