Quick clip: a photo comes alive in 5–15 seconds

“Quick clip” (Create menu) is a short video up to 15 seconds with one character from a photo: a person, cartoon character, pet or toy. For longer stories with scenes and several characters use “My films”. How it works: your photo is the first frame — the character, place and light come from it. The clip is built from 5-second scenes (5 sec = 1 scene, 10 = 2, 15 = 3), each continuing the previous one from its last frame. Each scene is animated by the LTX-2.5 model on our GPU server (LTX-2.3 as backup), then scenes are joined and voice, subtitles and music are added. Script — two modes, the “AI director” checkbox: • On (default): one field “What happens in the clip”. The AI (Gemini) splits the idea into 5-second scenes, invents the movements and, if the character speaks, short lines, and translates everything into English for the model. Handy when you just have an idea. • Off: a field appears for each scene. Your text goes to the model as written — no text AI is involved and the photo isn't described by AI either. The model understands English best — write simply and concretely: “the character waves and smiles at the camera”. If the character speaks, each scene has a line field voiced word for word (up to ~10 words per scene). Scene parameters (“Parameters” button): camera movement (static, push-in, pull-out, pan, orbit, follow, handheld), shot (wide, medium, close-up, detail), character movement intensity and emotion. They become precise wording for the model. Always on: same character, always fully in frame, same place, no other people and no scene changes — so the clip doesn't drift. Voice: “The character speaks” — choose a voice (ElevenLabs or Gemini), lips are synced with speech; “No speech” — movement only, no voice or subtitles. Background music — the “Background music” block (no music by default). Tips: a clearly visible character in the photo and simple actions give the most predictable result. The model follows your wishes but doesn't guarantee every detail — if a scene fails, simplify the text or change the parameters.