Wan 3.0 R2V Uncensored Video

Overview

Wan 3.0 R2V Uncensored by Wiro Partners generates short videos from a written prompt plus optional reference assets. You can feed it multiple images, short video clips, and short audio clips, then describe how each reference should influence the result. The model produces a finished MP4 clip, and it can include an audio track in the same generation. It’s useful when you need subject identity and visual style to stay stable across motion.

What you can build

• Consistent character clips using a small set of reference portraits and a scene prompt
• Style-locked videos for product shots, fashion looks, and creator visuals
• Motion matching by supplying a short reference clip and asking for a new setting
• Video variations for A/B testing while keeping the same subject references
• Prompt-driven edits where references define “who” and the prompt defines “what happens”

Inputs

• A written scene prompt that explains the shot, action, camera behavior, and style. You can point to references by index, like “Image 1” or “Video 1”.
• Optional reference images. You can provide up to 10 subject or style images.
• Optional reference videos. You can provide up to 5 clips. Keep the combined reference-video time within 15 seconds. Keep each clip under 100 MB.
• Optional reference audio. You can provide up to 5 audio clips. Keep the combined audio time within 15 seconds.
• A target video length in whole seconds. Choose any length from 2 to 30 seconds.
• An output quality tier. Choose 480p, 720p, or 1080p.
• An output aspect ratio. You can force common ratios (16:9, 9:16, 1:1, 4:3, 3:4) or let the model pick adaptively from your prompt and references.
• An option to include an audio track in the output. If you want sound, describe it in the prompt.
• An option to expand and rewrite your prompt for more detail. This can help when your prompt is short.
• An optional numeric seed for repeatable results. Use an integer from 0 to 2,147,483,647.

Outputs

• One generated MP4 video file at the chosen duration and resolution.
• The video file typically includes metadata such as width, height, frame rate, duration, and total frames.
• When audio is enabled, the MP4 includes an audio track generated with the visuals.
• The response may include the final seed that was used.
• The response may include the final prompt text after any prompt expansion.

Recommended settings

• If your prompt is under 1–2 sentences, turn on prompt expansion. It usually improves detail and adherence.
• If you need multiple takes with the same framing and subject, set a fixed seed and change only one prompt detail per run.
• If you enable audio, add explicit sound cues in the prompt. Name sources and timing (footsteps, room tone, music style).
• For identity consistency, use fewer, clearer references. Conflicting outfits and angles can blend.
• For reference videos, keep them short and focused on one motion pattern. Avoid rapid cuts and heavy compression.

Limitations

• Output duration is capped at 30 seconds per generation.
• Reference videos and reference audios are limited to 15 seconds total per request.
• Reference clips that are blurry, heavily compressed, or low frame rate can cause motion glitches.
• Mixed or contradictory references can cause identity drift, style blending, or sudden changes mid-clip.
• On-screen text can be inaccurate or unstable across frames.
• Audio texture can vary and may not match exact lyrics or precise phonemes.
• Low-quality inputs reduce control. This includes noisy images, scanned screenshots, collages, and messy background audio.

Safety & compliance

• This is an uncensored, mature-theme model. Restrict access to adult users where required.
• Don’t generate or request sexual content involving minors, or any exploitative content.
• Don’t upload faces, voices, or copyrighted media you don’t have rights to use.
• Don’t create deceptive impersonations of real people or misleading “real event” footage.
• Some deployments may still apply automated checks for illegal content. You are responsible for lawful use.