Video generation
99 video models across text-to-video, image-to-video, video-to-video editing, and clip extension.
Quick start
bash
# text-to-video
gen-ai generate -m seedance-2.0 -p "a fox running through autumn leaves" -d 8
# image-to-video
gen-ai generate -m kling-motion-control-v3 -p "slow push in" -i ./portrait.jpgjson
{ "name": "picsart_generate",
"arguments": { "model": "seedance-2.0", "prompt": "a fox running through autumn leaves", "duration": 8, "aspectRatio": "16:9" } }Input types
| Type | Command | Models |
|---|---|---|
t2v text→video | gen-ai generate | Seedance 2.5 / 2.0, Kling V3, Veo 3.1, Gemini Omni, Wan 3.0 / 3.0 Prime, Flux 3 Video, Hailuo 2.3, LTX, Luma Ray 3.2, HeyGen Video, HeyGen Video Avatar |
i2v image→video | gen-ai generate -i | Kling Motion Control, Wan 2.7 I2V, Runway Gen4 Ref, Picsart Effects Video, VEED Fabric, HeyGen, Creatify |
v2v video→video | gen-ai generate --video (or --video-urls) | Seedance Video Edit/Extend, Wan Video Edit, Runway Aleph, Grok Edit/Extend, LTX Extend/Retake |
Providers
| Provider | Models | Highlights |
|---|---|---|
| Seedance | Seedance 2.5 / 2.0 (+ Fast, Edit, Extend) | Reference image, keyframes, native audio; 2.5 supports 4–30s |
| Veo 3.1 / Fast / Lite, Gemini Omni | 1080p+, native audio; Omni adds 4K + video extension | |
| Kling | Kling 4.0, 4.0 Omni, V3, Video O1, Motion Control, Avatar, Effects | Up to 30s at 1080p; motion control from a still |
| Wan | Wan 3.0, Wan 3.0 Prime, Wan 2.7 (t2v/i2v/r2v/edit) | Versatile, crisp detail; Prime is up to 7x faster |
| Runway | Gen 4.5, Aleph, Gen4 Ref, Avatar | Editing & reference workflows |
| MiniMax | MiniMax H3, H3 Max (+ Turbo / Camera Controls), Hailuo 2.3 | Up to 2K; multimodal references |
| LTX | LTX 2.3 Pro/Fast, A2V, Extend, Retake | Audio-to-video & re-takes |
| Luma | Ray 3.2, Edit, Reframe | Reframing & fast generation |
| Picsart | Picsart Effects Video | Curated one-tap effect presets |
| Grok | Imagine Video, Edit, Extend | — |
| VEED · HeyGen · Creatify | Fabric, Talking Photo, Aurora | Avatar / talking-photo |
| ByteDance · OVI · Videography · HappyHorse | Video Enhance, OmniHuman 2.0, OVI, Videography, Happy Horse | Specialized |
| Meta · Depth Anything · Mirelo | SAM 3 Video, Video Depth Anything, Mirelo SFX 2.0 (Video to Video) | Segmentation, depth maps, synced soundtracks |
Extending clips
Use a dedicated extend model through gen-ai generate — Seedance, Grok, and LTX each have one:
bash
gen-ai generate -m seedance-2.0-video-extend -p "the camera keeps panning right" --video-urls ./clip.mp4Run gen-ai models info <id> to see which video flag an extend model takes (--video or --video-urls).
Common video parameters
| Param | CLI flag | Notes |
|---|---|---|
prompt | -p | Required (or stdin) |
aspectRatio | --ar | e.g. 16:9, 9:16, 1:1 |
resolution | -r | e.g. 720p, 1080p, 4k |
duration | -d | Seconds (model-dependent set) |
generateAudio | --audio-gen | Native audio track |
imageUrls / videoUrl | -i / --video | i2v / v2v inputs |