Sora 2: Text & Image to Video
OpenAI's Sora-2 and Sora-2 Pro video generation with synchronized audio.
About
This ComfyUI workflow demonstrates end-to-end text- and image-to-video generation using the OpenAIVideoSora2 node, with optional synchronized audio. You can start from a text prompt alone or condition the video with a reference image loaded via LoadImage. Because Sora-2 and Sora-2 Pro expect fixed input sizes, the workflow offers two branches to make your image fit: Option 1 uses ResizeAndPadImage to letterbox without cropping, and Option 2 crops to fill the frame. PreviewImage lets you confirm framing before you trigger the API call, and SaveVideo writes the returned video to disk.
What you provide
- Reading the workflow
Runs in ComfyUI. This workflow can't run on this page. Open it in Comfy Cloud with everything loaded, or download it for your own ComfyUI.
Try in CloudQuestions & answers
Use Sora-2 for 720x1280 or 1280x720 when you want faster, standard aspect results. Choose Sora-2 Pro if you need the additional sizes (1024x1792, 1792x1024) or higher fidelity. Always match your input image to the exact resolution you select.
Use Option1 - Pad image (ResizeAndPadImage) to letterbox and preserve the entire image, or use Option 2 - Crop image to fill the frame edge-to-edge. Check the result with PreviewImage before running the API. The API will reject mismatched sizes.
Yes. Bypass the image branch (select those nodes and press Ctrl-B) so OpenAIVideoSora2 receives only your text prompt. You can re-enable the image path later to switch back to image-to-video.
When available, the OpenAIVideoSora2 API returns a video that includes synchronized audio. SaveVideo writes the received video file to disk. If the API response is silent, your output will not contain audio.
Confirm you’re running in a secure environment that allows localhost/127.0.0.1, your API credentials are set in the OpenAIVideoSora2 node, and your firewall/proxy isn’t blocking the request. See the partner node overview and pricing links in the MarkdownNote.