Lesson Content

Handling common error conditions.

API errors include authentication failures (invalid or expired credentials), invalid request parameters (incorrect avatar ID, unsupported voice configuration), rate limit exceeded responses, generation failures (a generation request that starts but fails during processing), and credit/quota exhaustion. Each requires different handling – authentication and parameter errors require fixing the request; rate limits and transient service issues may be retried with backoff; quota exhaustion requires waiting or upgrading, not retrying.

Retry logic for transient failures.

For transient failures (rate limiting, temporary service unavailability), exponential backoff – increasing wait time between successive retries with a maximum retry count – is the appropriate pattern. Do not implement retry logic for non-retryable errors like quota exhaustion or invalid parameters.

import time

def generate_with_retry(payload, max_retries=3):
for attempt in range(max_retries):
response = make_generation_request(payload)
if response.status_code == 200:
return response.json()
if response.status_code == 429: # rate limited
time.sleep(2 ** attempt)
continue
raise Exception(f"Non-retryable error: {response.status_code}")
raise Exception("Max retries exceeded")

Rate limit and credit management.

Generation credits are consumed per video – automated pipelines can exhaust credits faster than manually-paced web interface use. Monitor credit consumption against the account's plan limits, and design pipelines with rate limiting in mind to prevent burst generation that exhausts credits unexpectedly.

Monitoring pipeline health.

Log generation request outcomes (success, failure type, generation time) for the pipeline. A rising failure rate, unusual error distribution, or generation time outliers are signals worth investigating – monitoring surfaces these before they become larger issues.