Omni Flash｜One line to a film
One photo or one sentence becomes a 5-10s cinematic short. A single picture is enough — shoot vertical for TikTok.
Gemini Omni Flash video generator: turn a photo or text into a 5-10s clip. Multimodal fusion, true physics, legible titles. Try it free online now!
Features
- Multimodal input——Character art, scene art, and text go in together. One face and one outfit held across the whole shot.
- True-to-physics motion——Gravity and inertia behave: splashes, sways, arcs and orbits read like footage, not morphing.
- Legible on-screen text——Headlines and label cards render sharp and readable — GLOW-level clear, synced with the scene.
- Single photo is enough——No training photos, no code. One phone picture is enough for a vertical clip.
How it works
- Upload or skip the photo Got a photo? It becomes the first frame. Got nothing? Pure text works too — the epic surfer shot used zero uploads.
- Describe the shot in words Say who moves, how, and how the camera travels — then pick length, framing, and look.
- Get a post-ready clip A 5-10s vertical clip, ready to post. Want longer? Ask the agent to continue the scene.
FAQ
- Can I do it with just one photo?
- Yes — one photo is its best trick. The photo becomes the first frame, so the face stays yours. Clear front-facing portraits work best; side profiles or sunglasses may drift, and then a second clear photo fixes it.
- How do I avoid fake faces and glitchy hands?
- Write small, real actions (blink, breathe, slight smile) instead of big stunts, and name the identity lock explicitly. Keep hands simple — holding a cup, resting naturally — with five natural fingers, and re-roll once if needed.
- Omni Flash vs Sora 2 vs Kling 3.0?
- It depends on the shot. Kling leads for realistic human motion and long takes, Hailuo for atmosphere, Vidu for multi-photo consistency — Omni Flash wins for fast drafts, fusion of several references, and readable on-screen text. Many creators draft here, then upscale hero shots elsewhere.
- Is it free? Do I need an app?
- Sign in and pay per generation — no subscription needed to start, no app download. It runs in your mobile browser; the vertical 9:16 option is made for phones.
- Can it do voiceover or a 40s video?
- One clip is 5-10 seconds by design. For a 40-second explainer, the agent generates connected segments and stitches them — tell it to continue and it extends the scene. Voiceover is added afterwards as narration, not a native audio track.
- Can I upload a video to edit?
- No. This tool takes images plus text only — there is no video-upload editing here, and no multi-round API-style edits. Each revision is a fresh generation that keeps everything you didn't ask to change.
Try it free