Media SVD (open-source video generation)

Context: $ARGUMENTS

Quick start

• Text-to-video: scripts/svd.py t2v --model ltx --prompt "..." --duration 5 --fps 24 --out out.mp4 → Step 3.
• Image-to-video: scripts/svd.py i2v --model cogvideox --init-image in.png --prompt "..." --duration 5 --out out.mp4 → Step 3.
• AnimateDiff image-model motion: scripts/svd.py animate --workflow animatediff.json --out out.mp4 → Step 4.
• Install a model's pip deps: scripts/svd.py install ltx → Step 2.
• Fetch weights (confirm required): scripts/svd.py download --model ltx --yes → Step 2.

When to use

• Generate a short video clip from a text prompt using permissively licensed open-source models only.
• Animate an existing still image into motion via image-to-video (i2v).
• Use AnimateDiff to add motion to any media-sd image model workflow.
• Need commercial clearance: all default models are Apache 2.0.
• Do NOT use for Stable Video Diffusion (SVD), Sora, Runway Gen-3, Kling — they are either non-commercial or proprietary APIs. HunyuanVideo has a commercial user-size cap (Tencent License) — mention but do not default to.