New idea
Submit text alone, text+image, or text+image+script. Each idea becomes one approvable card on the Create Videos tab.
Reference images optional
Drop one image for image-to-video, multiple images for per-image scripts (each clip will be stitched). Without an upload, the backend generates an image from your idea text.
Drop images here
or click to choose — up to 2 images per video
Add a script for each image below (or leave blank to use the main script).
Create Videos
loading…Generated videos
Combine clips into a reel
Click clips below in the order you want them to play. Each picked clip shows its position; click again to remove and the rest renumber. Pick 2–10 clips, choose a transition, hit Combine. The result appears on the Videos tab once ffmpeg finishes (usually 10–30s).
API keys & provider
Keys are stored in your browser cookies only. They are sent with every
request and used in-memory by the server. Empty fields fall back to
the server's .env values.
Phase 2 narration (TTS). Select a provider to enable per-scene voiceover. Leave Provider blank to keep videos narration-free (background music from Phase 1 still applies).
RunPod deployment. Cluster spins up an A6000 pod and reuses it across your queue (hourly billing). Serverless invokes a deployed endpoint per request (per-second billing, scale to zero, no warm-keep). Hugging Face Space runs against your configured Space endpoint. Per-idea overrides on the Compose form win over this default.