viral-video-replicator

To reverse-engineer a viral video, this Skill extracts frames using FFmpeg, transcribes audio via Seed-ASR-2.0, and analyzes frames with a Vision LLM to assemble structured Seedance 2.0 prompts for video replication. Upload your video and request replication to start.

Yes, you can swap faces or outfits when replicating a reference video using one of four material replacement modes: clone, face_swap, outfit_swap, or full_swap, allowing targeted element replacement within your generated Seedance 2.0 prompt.

Yes, FFmpeg is required to extract frames for video analysis. The Skill depends on FFmpeg and FFprobe to process reference videos, alongside ASR for audio transcription and a Vision LLM for structured visual analysis.

The best way to analyze video content for prompt generation is through structured frame extraction and audio transcription. This Skill combines FFmpeg frame extraction, Seed-ASR-2.0 transcription, and Vision LLM analysis to reverse-engineer videos into Seedance 2.0 prompts.

Yes, the video replication process supports batch processing, enabling you to analyze multiple reference videos and generate Seedance 2.0 prompts efficiently for replicating viral content at scale.

Speech transcription contributes to video prompt replication by using Seed-ASR-2.0 to extract audio dialogue and context, which the Vision LLM combines with extracted frames to assemble accurate Seedance 2.0 prompts for replicating the original viral video.