wan2.6-r2v is a video reference generation model in Alibaba's Wan 2.6 series, designed to bring character appearances from existing videos into new scenes and stories. It establishes character identity from reference material, then combines text descriptions to arrange actions, environments, and narratives. With multi-character and audio synchronization capabilities, it is suited for continuous character shorts, branded character content, and creative shot production.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
api.acedata.cloudwan2.6-r2vInput parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
Resolution, duration, frame rate, and format are native specifications; media submission and result retrieval use this platform's Wan video API.
The focus of video reference is character identity, rather than guessing a character's appearance again with text. wan2.6-r2v can extract character appearances from existing clips and use them to create new scenes. Prepare material in which the character is clearly identifiable, then describe the target environment, intended clothing, and actions. This is better suited for continuously creating content around a fixed character.
The model supports multiple characters and narratives, making it suitable for shorts depicting characters meeting, interacting, or completing actions together. When creating, specify each character's identity, position, and sequence of actions so that events within a short duration have a clear storyline. There is no need to fit an entire complex story into one clip; it can be split into segments by plot.
wan2.6-r2v's native capabilities include audio synchronization, making it suitable for planning sound requirements together with character actions and scene atmosphere. Prompts can clearly specify which events need sound accompaniment, rather than only listing visual tags. After production, check the correspondence between visuals and sound, then decide whether additional editing or voice-over is needed.
Using an existing character video as a reference, enter the location, actions, and emotional changes for this episode to generate new clips that preserve the character's identity. Suitable for creations such as having a character move from indoors to the street or enter different story environments. The deliverable is a short video that can continue to be edited, allowing the production team to review character traits and plot continuity clip by clip.
Design different usage scenarios, interaction methods, and opening actions around existing brand characters or on-camera talent, and use reference videos plus creative scripts to generate candidate shots. Suitable for first validating how characters perform in new scenes, then selecting appropriate clips for advertising production; product logos and text details should still be checked during the final production stage.
Organize a short plot into shot descriptions covering who appears, who acts, and what happens, then generate interaction clips with character reference materials. The deliverables can be used to evaluate composition, character relationships, and narrative pacing, helping the team discuss filming plans. Test key interactions first, then combine selected shots into a more complete story.
If the core task is to place a character from an existing video into a new scene, wan2.6-r2v better fits the workflow than describing the character from scratch. When you only have a text concept, you can choose wan2.6-t2v; when starting with a still image, consider wan2.6-i2v. Both have publicly available durations of 2–15 seconds, while this model supports 2–10 seconds, so consider both the material type and shot length when choosing.
wan2.6-r2v explicitly supports audio synchronization, multiple characters, and narrative capabilities, making it suitable for short films that need characters, events, and sound designed together. wan2.6-r2v-flash is a reference-video variant that emphasizes fast generation and can be considered when iteration speed matters. Do not assume that their sound controls are fully identical based solely on their names; test each one according to your actual creative goals.
Put the reference video in reference_video_urls, and describe the character and new scene in prompt; this model does not directly trim or edit the original video.
Specify model=wan2.6-r2v for /wan/videos, submit according to the reference video guide, and start with 5 seconds at 720P; Wan 2.6 does not use Wan 3's 30-second duration, automatic duration, or all-purpose media parameter.
Use async=true to save the task_id, query /wan/tasks or receive a callback; after completion, retrieve video_url, check the subject, actions, and sound, then proceed to editing.
Use the person in the reference video as the character, have them walk into a bookstore, pick up a book, and look at the camera, preserving their appearance, clothing, and vocal characteristics, with warm scene lighting.
Use reference_video_urls to submit identity references; check the character, scene transitions, and audiovisual performance, and do not treat the task as frame-by-frame editing of the original video.
It uses the character identity in a reference video as the basis for creation, applying the person's appearance to a new scene; image-to-video uses a static image as the generation starting point. If you already have a video character and want to continue telling their story, r2v is a better fit; if you mainly want to animate a single image, consider wan2.6-i2v.
The native output specifications of wan2.6-r2v are 2–10 seconds, with support for 720P and 1080P, and videos are 30 fps MP4 files. When designing scripts, arrange actions around short shots, and do not treat the 15-second limit of wan2.6-t2v and wan2.6-i2v as the output range for this model.
Submit model=wan2.6-r2v to POST /wan/videos, provide the reference video URL with reference_video_urls, and use prompt to describe the new scene. Set async=true to first obtain a task_id, then query the final status through /wan/tasks; after success, obtain the video URL for playback or download.
First explain the identity of the reference character in the new clip, then describe the environment, action sequence, and shot intent. For multi-character tasks, distinguish who does what, and avoid using only abstract terms such as “cinematic feel.” Keep the plot focused on one clear event, and check whether the generated character appearance and actions meet the creative goal.
wan2.6-r2v has native audio synchronization capabilities and can be used for short narratives with sound, but this does not make it an independent voice-over editing, precise lip-sync correction, or voice cloning tool. When creating, describe audio requirements together with scene events, and check the synchronization effect after completion; when precise dialogue or audio processing is needed, arrange post-production afterward.
Choose a development approach or application suited to this model and follow its setup guide.
↗Python SDK Integration Guide
SDK installation, authentication and examples↗TypeScript SDK Integration Guide
SDK installation, authentication and examples
↗Pricing
$1.28–$4.42· per generation
You pay$13,584for 142,800 credits · $0.0951 / credits46% OFF
You're billed in credits. Pick a top-up tier to see its per-credit rate — the larger the top-up, the cheaper each credit.
This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs