Describe a scene, a camera move or a mood, then choose duration, resolution and audio. The credit cost is shown before you generate.
Models that support it
Every video model available in Imagine AI, with supported scenes, durations, resolutions and credit pricing.
Long-form walkthroughs for prompting, choosing a model and finishing a clip.
Text to video, image to video and reference-driven shots with Seedance 2.5, Veo 3.1, Kling 4.0 and more. Start free — no credit card required.
Scenes, durations, resolutions, audio support and the lowest credit price for every model in the catalog.
Swipe sideways to see every model.
Start from the shot you need. Each scene page explains what the mode does, and the models below are the ones that can run it right now.
Describe a scene, a camera move or a mood, then choose duration, resolution and audio. The credit cost is shown before you generate.
Models that support it
Bring a still image to life. Upload the first frame, optionally a last frame, and choose how long the shot should run.
Models that support it
Upload reference images to keep a character, product or art style consistent, then describe the shot.
Models that support it
The same generator covers short ads, social clips, consistent characters and restyled footage. Pick the workflow that matches the job.
Shoot the product once, animate a still, and keep the label, shape and colour consistent across every cut. Useful for marketplace ads and social campaigns that need several versions quickly.
Describe the joke, the camera move or the reaction in text, then generate a short vertical clip. Most social formats only need 4–8 seconds, so a text-to-video shot is usually enough.
Reference-to-video keeps a face, a product or an art direction stable across shots. Upload the reference set once, then describe each new scene instead of rebuilding the character from scratch.
Video-to-video keeps the original motion and replaces the look. Start from footage you already own when the performance is right but the style, lighting or setting is not.
Start from the shot, the length and the budget — not from the model name.
Text-to-video is for shots that do not exist yet. Image-to-video animates a frame you already have. Reference-to-video is for characters or products that must stay consistent. Video-to-video restyles footage you already own. Choosing the wrong scene is the most common reason a result disappoints.
Durations range from 4 seconds to 30 seconds depending on the model, and not every model offers the same resolutions. The duration control only shows lengths the selected model supports, but the comparison table above lets you plan the edit before you spend credits.
Audio-capable models generate synchronized sound with the clip, which saves a separate sound pass for talking shots and ambience. Silent models are usually cheaper per second, so use them when you plan to add music or voice-over in the edit anyway.
Video is priced per second of output, and the rate changes with resolution, audio and quality tier. Generate a short test at the lowest settings first, confirm the motion and framing, then render the final version at the resolution you need.
Image-to-video and reference-to-video shots start from a still. These image tools make a clean frame before you animate it.
Each card opens the model's own page with its scenes, durations, resolutions and credit pricing.
Alibaba Wan 3.0 with native audio, 480p-1080p output and smart duration.
Kling 4.0 cinematic video with multi-shot support, 3-30 second clips at 720p or 1080p.
Google Veo 3.1 with synchronized audio, 4/6/8 second clips up to 4k.
Google Gemini Omni Flash 1.1: fast, cheap video with reference images and up to 4k.
MiniMax H3 video generation with reference images and up to 4K output.
ByteDance Seedance 2.5 with native audio, up to 30 second clips and 480p-1080p output.