Skip to main content
Automatically generate new video content from prompts, scripts, or topics using AI. AI Content Generation is designed for creators who want to produce video from scratch — no camera or footage required.

​
How It Works

The AI Content Generation tile takes your text prompt or topic and produces a complete video with visuals, narration, and structure. You define the subject matter, style, and format, and Mosaic handles the rest — generating scenes, selecting or creating visuals, and assembling the final output. AI Content Generation UI

​
Input & Settings

​
Prompt / Topic

Describe what the video should be about. Examples:
  • “5 tips for productivity while working from home”
  • “A 60-second explainer on how solar panels work”
  • “An Instagram Reel about the benefits of meditation”
Be specific about the subject, audience, and intended platform for best results.

​
Style

Choose the visual and tonal style of the generated content. Options may include:
  • Educational — clean, informative, structured
  • Social Media — fast-paced, bold, attention-grabbing
  • Cinematic — high production feel, dramatic pacing
  • Corporate — professional, polished, brand-safe

​
Duration

Set the target length of the generated video. Common options:
  • Short (15–30s) — TikTok, Reels, Shorts
  • Medium (60–90s) — Instagram, LinkedIn
  • Long (2–5 min) — YouTube, training content

​
Voice & Narration

Choose whether to include AI-generated narration and select a voice style. Options:
  • No narration — visuals and text only
  • Auto-narrate — AI writes and speaks the script
  • Custom script — you provide the narration text

​
Usage Recommendations

Use AI Content Generation to:
  • Rapidly produce social media content at scale
  • Create explainer videos from blog posts or articles
  • Generate video drafts for brainstorming and iteration
  • Build educational or training content without filming
AI Content Generation works great when combined with:
  • Captions (add subtitles for accessibility and engagement)
  • AI Music (add a background score)
  • Reframe (adapt output to different aspect ratios)
  • Destination (publish directly to social platforms)

​
API Info

  • Node ID: d898f2b1-3151-4231-a4d4-dd4d5a020b05

​
Node params

ParamTypeRequiredDefaultNotes
promptstringYes""Primary generation instruction.
model"kling-2.1-master" | "kling-1.6" | "helix" | "helix-fast" | "veo-3.1" | "veo-3.1-fast" | "veo-3" | "veo-3-fast" | "sora-2" | "sora-2-pro"NoUI default "kling-2.1-master"Video model selector.
aspect_ratio"16:9" | "9:16"No"16:9"Current UI restricts to these two ratios.
length_targetnumber (seconds)Nomodel-dependentModel-specific increments/ranges apply.
seed_image_uristringNounset”First frame” style/structure guide.
last_frame_imagestringConditional/model-limitedunsetSupported for Veo 3.1 paths; UI gates by model.
reference_imagesstring[]Conditional/model-limitedunsetUp to 3 reference images (UI).
sora_reference_imagestringConditional/model-limitedunsetSora-specific image guidance field.

​
Parameter groups

  • Generation core: prompt, model, aspect_ratio, length_target
  • Image guidance (model-specific): seed_image_uri, last_frame_image, reference_images, sora_reference_image

​
Model constraints

  • Veo 3/3.1 families: length_target uses 8-second increments, typically 8-64.
  • Kling families: length_target uses 5-second increments.
  • Sora families: planner path supports 4-second increments up to 60 seconds.
  • Legacy aliases helix and helix-fast are normalized to veo-3.1 and veo-3.1-fast.

​
Scenario requirements

  • Keep image-mode inputs mutually consistent:
    • Reference-images mode conflicts with first/last frame mode in UI.
    • Sora paths use sora_reference_image rather than seed_image_uri.

​
Example

{
  "prompt": "A cinematic product teaser with dramatic lighting and macro shots",
  "model": "veo-3.1",
  "aspect_ratio": "16:9",
  "length_target": 24,
  "seed_image_uri": "gs://bucket/seed-frame.png"
}