P-Video AI Video Generator
Generate videos from text prompts or images with P-Video AI. Up to 1080p at 48fps, with audio conditioning and prompt upsampling. 1–10 second clips.
Prompt
Create video from your photo (optional)
Upload a photo and we'll turn it into a video.
Video Settings
About P-Video
P-Video by Pruna AI generates videos from text prompts or images. It supports up to 1080p resolution at 48fps, with optional audio conditioning to sync visuals to a soundtrack. Use prompt upsampling to automatically enhance your descriptions, or draft mode for fast low-quality previews before committing to a full render.
Key capabilities
What sets it apart
Audio-driven generation. Upload a soundtrack and the model will sync the video's pacing and motion to your audio. Great for music videos, ads, and social content where timing matters.
Prompt upsampling. Turn a simple one-line description into a detailed cinematic prompt automatically. The model expands your text with camera angles, lighting, and style details to get better results.
Draft mode. Generate a quick low-quality preview to check composition and motion before running a full-resolution render. Saves time and credits during iteration.
High frame rate. Output at 48fps for buttery-smooth motion — ideal for slow-motion effects or fast-paced action sequences.
Input modes
Text-to-video — Describe what you want and the model generates it from scratch. Aspect ratio and duration are fully configurable.
Image-to-video — Upload a reference image and the model animates it into video. The aspect ratio is derived from the image automatically.
Audio-conditioned — Provide an audio file and the video will be generated to match its rhythm and length. Duration is determined by the audio.
Tips
- Keep prompt upsampling on for best results unless you want precise control over your prompt
- Use draft mode to iterate quickly, then switch it off for the final render
- When using audio, the video duration matches the audio length — the duration slider is ignored
- Aspect ratio is ignored when an input image is provided — the image dimensions are used instead