Generation Times & Performance

Processing time for lip sync generations depends on the model you choose, the duration of your input video, the video’s resolution, and current system load. Most generations complete within a few minutes. This page covers what to expect, what influences speed, how to monitor progress, and what to do when things take longer than expected.

Expected Generation Times

The table below shows typical generation times for each model. These are estimates based on standard input conditions and may vary during peak usage or with complex inputs.

What Affects Generation Speed

Several factors influence how long a generation takes to complete:

• Video duration — Longer videos require more frames to process. A 2-minute video takes roughly 4x longer than a 30-second video on the same model.
• Resolution — Higher resolution inputs require more compute. While all models process faces at 512x512, the overall pipeline still handles the full frame for composition.
• Model choice — lipsync-1.9.0-beta is the fastest model, designed for high-volume use. lipsync-2 balances speed and quality. lipsync-2-pro uses diffusion-based super resolution and is 1.5-2x slower than lipsync-2.
• System load — During peak usage periods, jobs may queue longer before processing begins. The PENDING status reflects time spent in the queue.
• Audio complexity — Audio with overlapping speakers, background noise, or unusual patterns may require additional processing cycles for accurate lip-to-audio alignment.

How to Check Generation Status

You can monitor the progress of a generation using any of these methods:

Polling

Call the GET /v2/generate/{id} endpoint at regular intervals. We recommend polling every 10 seconds.

1
import time
2
from sync import Sync
3
4
sync = Sync()
5
6
generation = sync.generations.get(job_id)
7
while generation.status not in ["COMPLETED", "FAILED", "REJECTED"]:
8
    time.sleep(10)
9
    generation = sync.generations.get(job_id)
10
11
print(f"Final status: {generation.status}")

Webhooks

Configure a webhook URL in your generation request to receive an HTTP POST notification when the job completes or fails. This is more efficient than polling for production workflows.

Studio

If you submitted the generation through Sync Studio, the current status is shown in the generations list. Refresh the page to see updated statuses.

Terminal Statuses

A generation ends in one of three terminal statuses:

Troubleshooting Slow Generations

If your generation is taking longer than expected, here are steps to diagnose and resolve the issue:

• Stuck in PENDING for more than 5 minutes — The processing queue may be busy during peak hours. Wait and monitor the status. Avoid resubmitting the same job, as this adds more jobs to the queue without canceling the original.
• Generation FAILED — Check the error response from the GET /v2/generate/{id} endpoint. The error and error_code fields describe what went wrong. See the Error Handling guide for a full list of error codes.
• Time-sensitive workloads — Use lipsync-1.9.0-beta for the fastest processing. It is 2-3x faster than lipsync-2 and costs less per second.
• Large batch processing — For high-volume jobs, use the Batch API to submit up to 500 generations in a single request with a 24-hour turnaround time.
• Check service status — Visit status.sync.so to see if there are any ongoing incidents or degraded performance affecting the platform.
• Visit the support KB — For additional troubleshooting, see Why is my generation taking so long or stuck in processing? in the support knowledge base.

Frequently Asked Questions