Next-generation cinematic video, now on Sunra
OpenAI's flagship image model — crisp text and scenes at 4K
Extreme aspect ratios — banners to ultra-wide and portraits
In-place video editing with synchronized audio
Alibaba's latest video model, now on Sunra
Photorealistic world simulation by OpenAI
Precision editing and character consistency
Image-to-video specialist with video-to-video editing
The agent analyzes your prompt intent and automatically decides whether to generate images, videos, or both. Say 'create a poster and animate it' and you get a still image and a video clip in one pass. Need batch output? Say '5 variations' and the agent splits into parallel tasks — up to 20 at once.
The agent selects from Nano Banana Pro, GPT Image 2, Nano Banana 2, and Nano Banana for images, and Seedance 2.0, Seedance 2.0 Fast, Kling O3, and Kling O1 for video. Each model's aspect ratio limits, resolution caps, and reference image counts are handled automatically — no manual configuration.
Use mark chips to tag regions on your canvas items — label 'the car' or 'the background' and the agent routes your images, videos, and audio to the correct pipeline slots. Supports start/end frames, reference images (up to 9 on Seedance, 4 on Kling), reference video, and audio references.
The agent auto-selects the optimal generation pipeline — text-to-video, image-to-video, reference-to-video, video-to-video, or video edit. Tell Kling 'make the car green' and it routes to in-place video editing. Same prompt on Seedance routes to video reference. When a model lacks support, the agent falls back gracefully.
Say '20 variations of this scene' and the agent splits your prompt into up to 20 concurrent generation tasks. Each task is independently routed to the optimal model and pipeline, and all results stream back to your canvas in real time. Batch workflows that used to take an afternoon finish in minutes.
The agent detects when your prompt implies speech, music, or ambient sound — 'a street musician playing guitar', 'a narrator introducing the product' — and automatically enables audio generation on supported models like Seedance 2.0. No manual audio toggle needed; the agent reads intent and configures the pipeline end to end.
Generate a product poster and a matching showcase video from one prompt. The agent assigns each output to the best-suited model automatically.
Say '5 different versions' or 'three variations with different styles'. The agent splits into parallel tasks and runs them concurrently — up to 20 outputs per prompt.
Upload reference images or video to the canvas. The agent auto-detects the right pipeline — image-to-video for stills, reference-to-video for style transfer, video-to-video for edits — with zero manual pipeline selection.
Run the same prompt across different models and compare side by side on canvas. Seedance for fast iteration, Kling for in-place video edits, GPT Image 2 for the best text rendering and brand accuracy.
Join thousands of creators, agencies, and brands who use Sunra every day.
The era of typing one prompt and hoping for the best is ending. AI agents now manage the entire video creation process, from script to final cut, across multiple specialized models.
With the recent launch of official agent execution environments, automating complex AI video tasks just became vastly more reliable for creators.
Build a repeatable AI video production workflow from scratch — covering planning, prompt writing, model selection, generation, editing, and delivery.
