Wan 3.0 Pro R2V Uncensored Video

Overview

Wan 3.0 Pro R2V Uncensored is a reference-to-video generator published by Wiro Partners. It conditions a Wan 3–class video model on your reference pack, then renders a new clip that keeps the subject and style consistent across frames.

You provide a prompt plus optional reference images, motion clips, and audio cues. The model returns a ready-to-use MP4 video at 1080p, 2K, or 4K. This is useful when you need the same person, outfit, or product to stay stable from start to finish.

What you can build

• Character-consistent clips for short narratives, loops, and scene variations
• Product shots that keep the same design while the camera moves
• Outfit and styling variations that still look like the same subject
• Motion transfer using a short reference video for pacing and camera language
• Audio-timed action beats when you want cuts and motion to match a rhythm
• High-resolution social exports for vertical, square, or landscape delivery

Inputs

• Reference images for subject identity or style. You can upload up to 10 images. Use common formats like JPG, PNG, WebP, or BMP.
• A required text prompt that describes the whole shot. You can point to your uploads by index, like “Image 1”, “Video 1”, and “Audio 1”.
• The target video length in whole seconds. It must be between 2 and 30 seconds.
• A quality tier for the final render. Choose 1080p, 2K, or 4K.
• An optional aspect ratio choice. You can force 16:9, 4:3, 1:1, 3:4, or 9:16, or let it adapt based on your prompt and references.
• Optional reference videos for motion or camera guidance. You can upload up to 5 clips, with 15 seconds total reference time. Each file must be 100 MB or smaller. Keep total input video time plus output time within 30 seconds.
• Optional reference audio for pacing and sound cues. You can upload up to 5 clips, with 15 seconds total reference time.
• An optional toggle to include an audio track in the generated output.
• An optional toggle that rewrites and expands short prompts into a longer, more detailed prompt.
• An optional seed number to make results more repeatable across runs.

Outputs

The model returns a single MP4 file.

The MP4 includes the rendered video stream at your selected resolution tier and aspect ratio. When audio output is enabled, it also includes an audio track in the same file.

Many implementations also return basic render metadata alongside the file, such as width, height, FPS, duration, and frame count.

Recommended settings

• If your prompt is short, turn on prompt expansion. It often improves detail and shot structure.
• If your prompt is already long and structured, leave prompt expansion off. This helps keep wording literal.
• For strong identity lock, use 3–6 reference images with different angles and lighting.
• If you add reference video, trim it to the shortest segment that shows the motion you want. This reduces rejection risk from duration limits.
• Set a seed when you need controlled iterations. Keep references and prompt text unchanged.

Limitations

• Output duration is capped at 30 seconds.
• Reference videos are capped at 5 files and 15 seconds total. Input video time plus output time must stay within 30 seconds.
• Reference audio is capped at 5 files and 15 seconds total.
• The model follows references by index. If your prompt points to the wrong index, results will drift.
• Very low-quality references can degrade results. Blurry faces, heavy compression, screenshots with text overlays, and collages often cause artifacts.
• High-resolution tiers can amplify temporal shimmer in fine textures like hair, lace, and detailed patterns.

Safety & compliance

• Only upload reference images, videos, and audio that you own or have rights to use.
• Don’t generate sexual content involving minors, or any content that exploits or endangers children.
• Don’t create non-consensual intimate imagery, harassment content, or deceptive impersonations of real people.
• Follow all applicable laws and the Wiro platform’s content rules, even when using an “uncensored” model.