Video Editing Nodes

Edit, modify, and transform existing videos.

Video Edit Gen4

AI-powered video editing with Runway Gen4 Aleph

Edits videos using natural language instructions via the Runway Gen4 Aleph model. This is a synchronous model -- it returns results directly without polling.

Property	Description
Inputs	video (video), ctx (text, optional prompt override)
Outputs	video (video)

Configuration:

Setting	Description
Aspect Ratio	Output dimensions (default: 16:9)
Prompt Override	Custom prompt (overrides flow context promptText)

How it works:

1. Validates video input
2. Builds payload with prompt and aspect ratio (uses promptOverride if set, otherwise ctx.promptText)
3. Calls wavespeedPredict directly (synchronous, no polling)
4. Returns edited video artifact

Example prompts:

"Convert to anime style"
"Apply cinematic color grading with teal and orange"
"Add rain effect to the scene"

---

Video Edit Kling

Kling O1-powered video editing

Edits videos using Kling's video-o1 editing model with optional image reference inputs.

Property	Description
Inputs	video (video), input (text, optional), image (image, optional reference)
Outputs	output (video)

Configuration:

Setting	Description
Model	kwaivgi/kling-video-o1/video-edit (default) or kwaivgi/kling-video-o1/video-edit-fast
Aspect Ratio	Output dimensions (default: 16:9)
Keep Original Sound	Preserve source audio (default: true)

Payload parameters:

• aspect_ratio -- Output aspect ratio
• images -- Optional reference image URLs for guided editing
• keep_original_sound -- Boolean to preserve original audio track
• prompt -- Edit instructions from flow context
• video -- Source video URL

Use cases:

• Style transfer while maintaining motion
• Object editing with image references
• Scene modifications with preserved audio

---

Video Swap

Swap elements between video and image

Replaces elements in a video using an image source. Supports two providers: PixVerse (object/subject swap) and Wan 2.2 (animation).

Property	Description
Inputs	video (video), image (image), input (text, optional prompt)
Outputs	video (video)

Providers:

Provider	Model Path	Modes	Description
PixVerse (default)	pixverse/swap	object	Object/element swap in video
Wan 2.2	wavespeed-ai/wan-2.2/animate	animate	Animate image using video motion

Configuration:

• Provider -- pixverse (default) or wan22
• Mode -- object (PixVerse) or animate (Wan 2.2)
• Resolution -- Output resolution (default: 720p)

---

Lip Sync

Synchronize audio with video of a speaking person

Makes a person in video appear to speak provided audio with accurate lip movements. Requires both video and audio inputs.

Property	Description
Inputs	video (video), audio (audio), ctx (text, optional -- InfiniteTalk only)
Outputs	video (video)

Available Models:

Model	Model Path	Features
Veed Lipsync	veed/lipsync	Basic lip sync (video + audio only)
Kling Lipsync	kwaivgi/kling-lipsync/audio-to-video	Optimized for Kling-generated videos
Sync Labs 2.0	sync/lipsync-2	Configurable sync mode
Sync Labs 2.0 Pro	sync/lipsync-2-pro	Premium quality with sync modes
InfiniteTalk	wavespeed-ai/infinitetalk/video-to-video	Prompt-guided, configurable resolution and seed
InfiniteTalk Fast	wavespeed-ai/infinitetalk-fast/video-to-video	Faster generation with prompt support

Sync Modes (Sync Labs only):

Mode	Description
cut_off	Cut video at audio length (default)
loop	Loop video to match audio length
bounce	Play forward then backward
silence	Add silence to match video
remap	Remap video timing to audio

InfiniteTalk-specific settings:

• Prompt -- Text guidance via the ctx input handle
• Resolution -- 480p (default) or 720p
• Seed -- Reproducibility (-1 for random)

How it works:

1. Validates both video and audio inputs
2. Extracts optional prompt from connected text node (for InfiniteTalk models)
3. Builds payload via buildLipSyncPayload based on selected model
4. Submits to generate-image edge function
5. Polls poll-wavespeed-job for up to 6 minutes

Example workflow:

Text to Audio --> Lip Sync
                     ^
               Upload Video

Use cases:

• Dubbing videos to different languages
• Creating talking avatars
• AI-generated character dialogue
• Voiceover synchronization

caution

Use responsibly. Only modify videos where you have appropriate permissions.

---

Video Editing Workflow Examples

Style Transfer Pipeline

Upload Video --> Video Edit Gen4 --> Library
                      |
              "Cinematic color grading
               with teal and orange"

Kling Edit with Reference

Upload Video --> Video Edit Kling --> Library
Upload Image -----^
                   |
           "Match the style of the reference image"

Lip Sync Dubbing

Text to Audio --> Lip Sync --> Library
                     ^
              Upload Video

---

Tips for Video Editing

Quality Preservation

1. Start with the highest quality source
2. Minimize the number of edit passes
3. Use appropriate resolution for the final output

Prompt Tips

Be specific about what to change:

"Change the sky to a dramatic sunset with orange
 and purple clouds, keep the foreground unchanged"

Include what to preserve:

"Apply vintage film look while preserving
 skin tones and maintaining sharpness on faces"

Lip Sync Best Practices

1. Face visibility -- Face should be clearly visible and well-lit
2. Audio quality -- Clean, high-quality audio produces better sync
3. Duration matching -- Choose appropriate sync mode for audio/video length differences
4. Model selection -- Use Sync Labs Pro for highest quality, InfiniteTalk for prompt-guided control