PixVerse / reference-to-video-v6

## Overview Reference-to-Video V6 by PixVerse turns a small set of reference images into a guided video clip. You describe the action in text and point to each image inside the prompt. The model uses those images to keep key subjects and style consistent while it creates new motion and camera changes. It returns a single MP4 video you can download, with optional generated audio. ## What you can build - Product shots that keep the same item design across a moving scene - Character clips based on a concept image, with new actions and framing - Brand style videos that keep a fixed look from a reference frame - Before/after style reveals driven by one reference and a motion prompt - Social shorts built around a single hero image and a clear action beat ## Inputs - 1–3 reference images you want the video to follow. Use clear JPG, PNG, or WEBP files. Keep the main subject large in frame. - A text prompt that describes the scene and motion. Mention each uploaded image in the prompt with tokens like @ref1, @ref2, and @ref3. - A target clip length. This model run offers 5, 10, or 15 seconds. - A target output quality setting: 360p, 540p, 720p, or 1080p. - An aspect ratio for the frame. Options include 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, and 21:9. - An optional seed number if you want repeatable results from the same inputs. - An optional switch to generate audio together with the video. ## Outputs - One downloadable MP4 video file. - The video matches the chosen resolution, aspect ratio, and duration. - If audio is enabled, the MP4 includes a synced audio track that can contain music, sound effects, or spoken dialogue. ## Limitations - You can only provide up to 3 reference images for a single run. - Very long prompts can fail. PixVerse platform limits prompts to 2,048 characters. - Image uploads can be rejected if they exceed platform limits. PixVerse platform caps images at 20 MB and a 10,000 px longest side. - Content moderation can block a request. This can happen for the text prompt or the reference images. - Low-quality inputs can reduce consistency. Heavy compression, tiny subjects, cluttered backgrounds, or text-heavy images can cause drift. - The prompt guides motion and framing, but exact poses and camera moves aren’t guaranteed. ## Safety & compliance - Don’t generate hate speech or harassment. - Don’t generate sexually explicit content. - Don’t generate excessive violence, gore, or disturbing content. - Don’t generate deliberate misinformation. - Don’t include private personal data about real people. - Don’t infringe copyright or trademark rights. Only use references you own or have permission to use. - Don’t request or promote illegal activity.