Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.together.ai/llms.txt

Use this file to discover all available pages before exploring further.

Using a coding agent? Install the together-video skill to let your agent write correct video generation code automatically. See agent skills for details.

​
Generate a video

Video generation is asynchronous: you create a job, receive a job ID, and poll for completion. 
import time
from together import Together

client = Together()

# Create a video generation job
job = client.videos.create(
    prompt="A serene sunset over the ocean with gentle waves",
    model="minimax/video-01-director",
    width=1366,
    height=768,
)

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

# Poll until completion
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("Video generation failed")
        break

    # Wait before checking again
    time.sleep(60)
Example output when the job is complete: 
{
  "id": "019a0068-794a-7213-90f6-cc4eb62e3da7",
  "model": "minimax/video-01-director",
  "status": "completed",
  "size": "1366x768",
  "seconds": "6",
  "outputs": {
    "cost": 0.28,
    "video_url": "https://api.together.ai/shrt/DwlaBdSakNRFlBxN"
  },
  "created_at": "2025-10-20T06:57:18.154804Z",
  "completed_at": "2025-10-20T07:00:12.234472Z"
}
When a job fails, the response includes an error object instead of outputs: 
{
  "id": "019a0068-794a-7213-90f6-cc4eb62e3da7",
  "model": "minimax/hailuo-02",
  "status": "failed",
  "error": {
    "message": "Unsupported use of 'negativePrompt' parameter. ...",
    "code": "unsupportedParameter"
  },
  "outputs": null
}
Job status reference:
StatusDescription
queuedJob is waiting in queue.
in_progressVideo is being generated.
completedGeneration successful, video available.
failedGeneration failed, check error.message.
cancelledJob was cancelled.

​
Supported models

For the current list of video models, including duration, resolution, FPS, and keyframe support per model, see the serverless catalog or the dedicated endpoint model catalog.

​
Troubleshooting

​
Video doesn’t match prompt well

  • Increase guidance_scale to 8-10.
  • Make prompt more descriptive and specific.
  • Add negative_prompt to exclude unwanted elements.

​
Video has artifacts

  • Reduce guidance_scale (keep below 12).
  • Increase steps to 30-40.
  • Adjust fps if motion looks unnatural.

​
Generation is too slow

  • Reduce steps (try 10-20 for testing).
  • Use shorter seconds during development.
  • Lower fps for slower-paced scenes.

​
URLs expire

  • Download videos immediately after completion.
  • Don’t rely on URLs for long-term storage.

​
Next steps