Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/UnkleFunk/HouseMusicSwarm-/llms.txt

Use this file to discover all available pages before exploring further.

The Video Generation Agent is a full-stack video production specialist that creates, edits, and assembles AI-generated video clips. Before any generation, it gathers five key inputs from the user — length, format, style, sound, and audience — then builds model-optimized prompts of 100 words or more and selects the best available provider. All finished videos are stitched into a final cut immediately with CombineVideos, and paths are always included in the response.

​
Generation tools

GenerateVideo

Core text-to-video and image-to-video generation tool. Supports three providers: Veo (default), Sora, and Seedance 1.5 Pro. Accepts optional first_frame_ref (image-to-video for all models) and asset_image_ref (Veo-only, keeps a subject consistent while allowing camera movement). Returns the video path, thumbnail spritesheet, and last frame for chaining.

CombineVideos

Merges an ordered list of video clips into a single MP4 using instant-cut transitions (ffmpeg concat demuxer). Used automatically after all segments are generated to deliver a final cut without waiting for per-clip approval.

TrimVideo

Removes unwanted seconds from the start or end of a clip using ffmpeg. Essential for cleaning up motion artifacts at clip boundaries or adjusting exact duration before combining.

​
Editing tools

EditVideoContent

AI-powered video transformation with three modes:
  • edit — transforms content via prompt using Kling O3 Standard (fal.ai), preserving motion. Best for targeted content changes with small composition shifts. Max ~10s; trim longer clips first.
  • remix — re-generates a Sora video job with a new prompt at higher fidelity. Composition and pacing can change significantly.
  • extend — appends 8 seconds of continuation to a Veo video. Requires the original veo_video_uri. Currently limited to 16:9 source videos.

AddSubtitles

Burns animated, word-level subtitles into a video. Uses OpenAI Whisper API for automatic transcription and precise timing. Supports configurable font size, position (top/center/bottom), words per clip, and highlight color. Accepts the original script as a transcription hint.

EditAudio

Replaces a video’s visuals with another video’s visuals while keeping the original audio track, or vice versa. Useful for laying b-roll footage over a narration track. Supports pad_seconds for timing offsets between audio and visual sources.

EditImage

Edits a single video frame or reference image using Gemini 2.5 Flash Image. Accepts a natural language edit prompt and produces up to 4 variants. Used for compositing reference frames before image-to-video generation.

GenerateImage

Creates a reference image using Google’s Gemini 2.5 Flash Image model. Use this to generate studio product shots, character concepts, stylized art, or any visual asset that will serve as a first_frame_ref or asset_image_ref for subsequent video generation. Outputs are saved to mnt/{product_name}/generated_images/.

CombineImages

Merges multiple images into a single composition according to a text instruction using Gemini 2.5 Flash Image. Use this to overlay a logo onto a product shot, composite a subject onto a background, or build any exact scene before animating it with GenerateVideo. Requires at least two image references.

​
Supported providers

Requires GOOGLE_API_KEYDefault provider for most video generation tasks.
  • Model: veo-3.1-generate-preview (use veo-3.1-fast-generate-preview when speed is priority)
  • Durations: 4s, 6s, 8s
  • Aspect ratios: 16:9 and 9:16
  • Audio: Explicit prompt-driven audio generation including dialogue (quoted text), sound effects, and ambient noise
  • Unique feature: asset_image_ref — feed a subject/product image to keep it visually consistent while the camera moves freely around it
  • Extension: Veo clips can be extended by 8 seconds using EditVideoContent with action="extend"

​
Example prompts

  • “Generate a product launch video with smooth camera movement, 24 seconds total”
  • “Add subtitles to this video in English”
  • “Combine these clips into a 60-second reel”
  • “Create an animated explainer about how our AI works — cinematic, 16:9, no voiceover”
  • “Extend this Veo clip by 8 more seconds, same scene”
  • “Replace the background music in this clip with the audio from another track”
Video generation requires at least one provider key (OPENAI_API_KEY, GOOGLE_API_KEY, or FAL_KEY). If a requested model is unavailable, the agent switches to an available alternative and informs you. Veo is the recommended default — it offers the best combination of quality, speed, and explicit audio control.

Build docs developers (and LLMs) love

Get started for freeTalk to us