Video Overview - PraisonAI

Default model: When you omit llm=/--model/model=, every surface — SDK VideoAgent, capability video_generate, MCP praisonai.videos.generate, and CLI praisonai videos — falls back to openai/sora-2 (the canonical default from VideoAgent.DEFAULT_MODEL).

Generate Video

Basic
Advanced

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="openai/sora-2")
video = agent.generate("A cat playing with yarn")
print(video.id)

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="openai/sora-2")

# Generate
video = agent.generate("A sunset timelapse", seconds="8", size="1920x1080")

# Wait and download
completed = agent.wait_for_completion(video.id)
if completed.status == "completed":
    agent.download(video.id, "sunset.mp4")

# List all videos
videos = agent.list()

Providers

OpenAI Sora

Sora-2

Gemini

Omni Flash (default), Veo 3.1

RunwayML

Gen-4

Azure

Sora

OpenRouter

Veo, Kling, Hailuo, Seedance, Wan…

Supported models

Only models with a LiteLLM video-generation backend work in VideoAgent. Pick a model that is in LiteLLM’s chat/cost catalogue but has no video route (for example xai/grok-imagine-video-*) and VideoAgent raises a clear ValueError naming models that do work. Known-good defaults:

Model	Provider	Notes
`openai/sora-2`	OpenAI	Default when `llm=` is omitted
`gemini/veo-3.1-generate-preview`	Google	Also `gemini/veo-3.1-lite-generate-preview`
`runwayml/gen4_turbo`	RunwayML	Image-to-video

Chat/cost catalogue membership is not enough — the model must have a provider-specific video config in LiteLLM. If LiteLLM later adds a video backend for your provider, VideoAgent picks it up automatically; nothing is hardcoded to a denylist.

Error you’ll see on an unsupported model

ValueError: 'xai/grok-imagine-video-1.5' has no LiteLLM video-generation backend
(it may be listed for chat pricing only). Use a supported video model, e.g.:
openai/sora-2, gemini/veo-3.1-generate-preview, runwayml/gen4_turbo

Switch llm= to a supported model and re-run:

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="openai/sora-2")  # instead of xai/grok-imagine-video-1.5
video = agent.generate("A cat playing with yarn")

Desktop app

Prefer a native app? The Desktop Video tab generates an MP4 from a prompt with a built-in model picker (MiniMax, Sora 2, Veo 3.1) — no code required.

Desktop → Video

Generate video from the Desktop app

Troubleshooting

ValueError: '...' has no LiteLLM video-generation backend

The model you passed to llm= is in LiteLLM’s model/cost catalogue but has no video-generation provider config. This is common for xAI Grok Imagine ids (xai/grok-imagine-video-*), which are chat-priced only.Fix: switch to one of the supported models from the error message:

from praisonaiagents import VideoAgent

agent = VideoAgent(llm="openai/sora-2")  # instead of xai/grok-imagine-video-1.5

If LiteLLM later adds a video backend for your preferred provider, VideoAgent picks it up automatically — nothing is hardcoded to a denylist.

AuthenticationError

The model’s video backend does exist, but your API key is missing or wrong. This error propagates unchanged — VideoAgent never masks it. Set the provider’s env var (e.g. OPENAI_API_KEY, GEMINI_API_KEY, RUNWAYML_API_SECRET) and re-run.

video size is not supported

The working backend rejected a specific request parameter (size, duration, prompt format). This is a bad request to a real video backend — not a missing backend — so it propagates unchanged. Fix the request, not the model: for example, drop size= to use the provider default.

​Generate Video

​Providers

OpenAI Sora

Gemini

RunwayML

Azure

OpenRouter

​Supported models

​Error you’ll see on an unsupported model

​Desktop app

Desktop → Video

​Troubleshooting

Generate Video

Providers

Supported models

Error you’ll see on an unsupported model

Desktop app

Troubleshooting