Skip to main content

​
Wan 2.7

Wan 2.7 is a family of video generation models supporting text-to-video, image-to-video with keyframe control, reference-based character/object consistency, and video editing. All models output 720P or 1080P video at 30fps in MP4 format.
ModelAPI StringBest ForDuration
Wan 2.7 T2VWan-AI/wan2.7-t2vText-to-video with audioUp to 15s
Wan 2.7 I2VWan-AI/wan2.7-i2vImage-to-video, keyframe control, video continuationUp to 15s
Wan 2.7 R2VWan-AI/wan2.7-r2vCharacter/object consistency from reference images or videosUp to 10s
Wan 2.7 Video EditWan-AI/wan2.7-videoeditInstruction-based editing, style transferUp to 10s

​
Text-to-Video

Generate a video from a text prompt. Video generation is asynchronous — you create a job, receive a job ID, and poll for the result. 
import time
from together import Together

client = Together()

job = client.videos.create(
    prompt="A small cute cartoon kitten general in golden armor stands on a cliff, commanding an army of mice charging below. Epic ancient war atmosphere, dramatic clouds over snowy mountains.",
    model="Wan-AI/wan2.7-t2v",
    resolution="720P",
    ratio="16:9",
    seconds="10",
)

print(f"Job ID: {job.id}")

while True:
    status = client.videos.retrieve(job.id)
    print(f"Status: {status.status}")

    if status.status == "completed":
        print(f"Video URL: {status.outputs.video_url}")
        break
    elif status.status == "failed":
        print(f"Error: {status.error}")
        break

    time.sleep(60)

​
Text-to-Video with Audio

Drive video generation with an audio file using media.audio_inputs. The model synchronizes the generated video to the audio — useful for lip sync, beat-matched motion, or narration-driven scenes. If no audio is provided, the model automatically generates matching background music or sound effects. 
import time
from together import Together

client = Together()

job = client.videos.create(
    prompt="A graffiti character comes to life off a concrete wall, rapping energetically under an urban railway bridge at night, lit by a lone streetlamp.",
    model="Wan-AI/wan2.7-t2v",
    resolution="720P",
    ratio="16:9",
    seconds="10",
    media={
        "audio_inputs": [
            "https://example.com/rap-audio.mp3",
        ],
    },
)

print(f"Job ID: {job.id}")

while True:
    status = client.videos.retrieve(job.id)
    print(f"Status: {status.status}")

    if status.status == "completed":
        print(f"Video URL: {status.outputs.video_url}")
        break
    elif status.status == "failed":
        print(f"Error: {status.error}")
        break

    time.sleep(60)
Audio constraints: WAV or MP3 format, 3-30 seconds, up to 15 MB. If the audio is longer than the video duration, it will be truncated. If shorter, the remaining portion of the video will be silent.

​
Image-to-Video

Animate a still image by using it as the first frame. Pass images via media.frame_images with frame set to "first" or "last". 
import time
from together import Together

client = Together()

job = client.videos.create(
    prompt="A black cat curiously gazes up at the sky. The camera slowly rises from eye level to a bird's-eye view, capturing the cat's curious eyes.",
    model="Wan-AI/wan2.7-i2v",
    resolution="720P",
    ratio="16:9",
    seconds="5",
    media={
        "frame_images": [
            {
                "input_image": "https://example.com/cat.png",
                "frame": "first",
            }
        ],
    },
)

print(f"Job ID: {job.id}")

while True:
    status = client.videos.retrieve(job.id)
    print(f"Status: {status.status}")

    if status.status == "completed":
        print(f"Video URL: {status.outputs.video_url}")
        break
    elif status.status == "failed":
        print(f"Error: {status.error}")
        break

    time.sleep(60)

​
First and Last Frame Control

Provide both a starting and ending frame to control the video’s transition. The model generates smooth motion between the two keyframes. 
import time
from together import Together

client = Together()

job = client.videos.create(
    prompt="Smooth cinematic transition with natural motion",
    model="Wan-AI/wan2.7-i2v",
    resolution="720P",
    ratio="16:9",
    seconds="5",
    media={
        "frame_images": [
            {"input_image": "https://example.com/start.png", "frame": "first"},
            {"input_image": "https://example.com/end.png", "frame": "last"},
        ],
    },
)

print(f"Job ID: {job.id}")

while True:
    status = client.videos.retrieve(job.id)
    print(f"Status: {status.status}")

    if status.status == "completed":
        print(f"Video URL: {status.outputs.video_url}")
        break
    elif status.status == "failed":
        print(f"Error: {status.error}")
        break

    time.sleep(60)

​
Video Continuation

Continue from an existing video clip using media.frame_videos. The model generates new content that seamlessly extends the input video. 
import time
from together import Together

client = Together()

job = client.videos.create(
    prompt="A dog wearing sunglasses skateboarding down a street, 3D cartoon style.",
    model="Wan-AI/wan2.7-i2v",
    resolution="720P",
    ratio="16:9",
    seconds="15",
    media={
        "frame_videos": [
            {"video": "https://example.com/skateboarding-clip.mp4"},
        ],
    },
)

print(f"Job ID: {job.id}")

while True:
    status = client.videos.retrieve(job.id)
    print(f"Status: {status.status}")

    if status.status == "completed":
        print(f"Video URL: {status.outputs.video_url}")
        break
    elif status.status == "failed":
        print(f"Error: {status.error}")
        break

    time.sleep(60)

​
Reference-to-Video

Generate video featuring a specific person or object by providing reference images or videos via media.reference_images or media.reference_videos. The model maintains the character’s appearance throughout the generated video. Multiple references can be passed for multi-character scenes. 
import time
from together import Together

client = Together()

job = client.videos.create(
    prompt="A person dancing on stage",
    model="Wan-AI/wan2.7-r2v",
    resolution="1080P",
    ratio="16:9",
    seconds="5",
    media={
        "reference_videos": [
            {"video": "https://example.com/character-reference.mp4"},
        ],
    },
)

print(f"Job ID: {job.id}")

while True:
    status = client.videos.retrieve(job.id)
    print(f"Status: {status.status}")

    if status.status == "completed":
        print(f"Video URL: {status.outputs.video_url}")
        break
    elif status.status == "failed":
        print(f"Error: {status.error}")
        break

    time.sleep(60)

​
Video Editing

Edit an existing video with text instructions using media.source_video. Optionally pass media.reference_images to guide the edit with a visual reference. 
import time
from together import Together

client = Together()

job = client.videos.create(
    prompt="Replace the background with the ocean",
    model="Wan-AI/wan2.7-videoedit",
    resolution="720P",
    ratio="16:9",
    media={
        "source_video": "https://example.com/input-video.mp4",
    },
)

print(f"Job ID: {job.id}")

while True:
    status = client.videos.retrieve(job.id)
    print(f"Status: {status.status}")

    if status.status == "completed":
        print(f"Video URL: {status.outputs.video_url}")
        break
    elif status.status == "failed":
        print(f"Error: {status.error}")
        break

    time.sleep(60)

​
Parameters

ParameterTypeDescriptionDefault
promptstringText description of the video to generate (up to 5,000 characters)Required
modelstringModel identifier (see model table above)Required
resolutionstringVideo resolution tier (720P, 1080P)"1080P"
ratiostringAspect ratio (16:9, 9:16, 1:1, 4:3, 3:4)"16:9"
secondsstringVideo duration in seconds. T2V and I2V: 2-15. R2V and Video Edit: 2-10."5"
seedintegerRandom seed for reproducibility (0-2,147,483,647)Random
negative_promptstringElements to exclude from generation (up to 500 characters)-
mediaobjectMedia inputs for the request (see schema and compatibility below)-

​
Media Object

The media object is the unified way to pass images, videos, and audio into video generation requests. 
{
  "prompt": "...",
  "model": "...",
  "media": {
    "frame_images": [],
    "frame_videos": [],
    "reference_images": [],
    "reference_videos": [],
    "source_video": "",
    "audio_inputs": []
  }
}
FieldTypeDescription
frame_imagesarrayKeyframe images for I2V. Each item: {input_image, frame} where frame is "first" or "last".
frame_videosarrayInput video clips for video continuation (I2V). Each item: {video: "url"}.
reference_imagesarrayReference images for character/object consistency (R2V) or visual guidance (Video Edit).
reference_videosarrayReference videos for character/object consistency (R2V). Each item: {video: "url"}.
source_videostringSource video URL to edit (Video Edit).
audio_inputsarrayAudio file URLs to drive generation — lip sync, beat-matched motion, etc. (T2V, I2V). Each item: "url". WAV or MP3, 3-30s, up to 15 MB.

​
Media Compatibility by Model

Not all media fields are supported on every model. Unsupported fields are rejected.
media fieldT2VI2VR2VVideo Edit
frame_images-First and/or last frame--
frame_videos-Single video clip--
reference_images--MultipleSingle
reference_videos--Multiple-
source_video---Single (required)
audio_inputsSingleSingle--

​
Prompting Tips

Wan 2.7 supports both Chinese and English prompts. Detailed, descriptive prompts produce the best results — include subject, action, style, camera movement, and atmosphere.
Write descriptive prompts. Instead of “a cat walking,” try “A small black cat walks gracefully through a sunlit garden, soft bokeh background, gentle breeze rustling the flowers, cinematic slow motion.” Use negative prompts to avoid common artifacts: 
low resolution, errors, worst quality, low quality, incomplete, extra fingers, bad proportions, blurry, distorted
Control aspect ratio and resolution. Use resolution and ratio to set output dimensions:
Aspect Ratio720P Dimensions1080P Dimensions
16:91280x7201920x1080
9:16720x12801080x1920
1:1960x9601440x1440
4:31104x8321648x1248
3:4832x11041248x1648

​
Next Steps