Generate video from text
Describe the scene, action, camera, lighting, and pacing when you want the model to build the entire shot.
Browse by modality
Video
Generate video from text, images, and references.
What you can do
Video models can generate a scene from text, animate a still image, follow reference materials, or transform an existing clip. They differ in motion quality, control over characters and shots, native sound, duration, speed, and moderation.
Animate a photo or illustration
Use a starting image to retain the composition while adding camera movement, character motion, or environmental effects.
Control a scene with references
Provide first and last frames, character or style references, audio, or a motion clip when the selected model supports them.
Edit, translate, or add a presenter
Specialized workflows can revise a clip from instructions, translate speech with lip sync, or create a talking presenter.
Inputs
- A prompt describing the shot, action, camera, and sound
- A starting image, and in supported workflows a final frame
- Optional image, video, motion, or audio references supported by the selected model
- Duration, aspect ratio, resolution, and other model-specific controls
Outputs
- A generated or transformed video clip
- Native sound only in models that explicitly support audio generation
- Model-specific duration and resolution rather than one shared format for the entire catalog
How to choose a model
Choose by workflow first: text, a starting image, several references, an existing video, or native audio. Then compare speed, quality, supported duration, and cost inside the relevant family.
Seedance 2
Seedance 2 covers text, frames, multimodal references, and native audio; its LR variants are useful when false prompt rejections are a concern.
View model details →
Kling Omni Video
Kling 3.0 Omni combines generation, references, video editing, native sound, and multi-shot prompting in one model.
View model details →
Veo 3.1
Veo 3.1 offers separate workflows for text, starting images, reference images, first-to-last-frame interpolation, and faster iterations.
View model details →
Hailuo Video
Hailuo is a straightforward option for short text-to-video and image-to-video clips, with standard and faster variants.
View model details →
Reviewed examples
The examples below are reviewed Omnipix generations. Open the family pages to see more outputs and the controls available for each version.
3.0 Omni
View model details →Limitations and trade-offs
- Character identity, object shape, and fine details can drift during motion or between shots.
- Native audio, reference video, first and last frames, and video editing are model-specific capabilities.
- Longer, higher-resolution, and reference-heavy generations generally take more time and credits.
- Moderation rules differ between providers and may reject a prompt that another model accepts.
All available model families
Alibaba
View detailsBlack Forest Labs
View detailsByteDance
View detailsSeedance 1.5
Seedance 1.5 Pro jointly generates picture and sound from text or an opening image, with optional last-frame guidance.
Seedance 2
Seedance 2.5 is the flagship text and multimodal-reference model, while Seedance 2.0 offers standard, Fast, Mini, and LR alternatives.
RecommendedStandard Seedance may refuse harmless prompts and synthetic people. Start with LR for substantially fewer moderation false positives; choose Full LR, Fast LR, or Mini LR for the quality and speed you need.
HeyGen
View detailsKling AI
View detailsKling 2.5 Turbo
Kling 2.5 Turbo provides text-to-video and image-to-video variants in Standard and Pro configurations.
Kling Avatar v2
Kling Avatar v2 turns one portrait and a speech recording into a synchronized talking-avatar video.
Kling Motion Control
Kling Motion Control transfers the movement from a reference video to the character or scene shown in a reference image.
Kling O1
Kling O1 combines text, first and last frames, image references, video references, and natural-language video editing in one model.
Kling Omni Video
Kling Video 3.0 Omni unifies text, image, reference, and video-editing workflows with optional native audio and multi-shot prompting.