Troubleshooting¶

Common issues and solutions.

Authentication Issues¶

"Invalid API key"¶

Problem: API key is not valid or expired

Solutions:

1. Check key prefix: AIzaSy... for Gemini API
2. Regenerate key at Google AI Studio
3. Ensure key is not quoted incorrectly

# Check key format
echo $GOOGLE_API_KEY | head -c 10
# Should show: AIzaSy...

"Permission denied" / "403 Forbidden"¶

Problem: API access not authorized

Solutions:

1. Enable Vertex AI API in Cloud Console
2. Check billing is enabled
3. Verify API key has correct permissions

# For Vertex AI, use OAuth instead of API key
gcloud auth application-default login
TOKEN=$(gcloud auth application-default print-access-token)
python video_gen.py "test" --key "$TOKEN"

"401 Unauthorized"¶

Problem: Authentication failed

Solutions:

1. Vertex AI keys (AQ...) need OAuth for some operations
2. Token may be expired - refresh it

# Refresh OAuth token
gcloud auth application-default login

Video Generation Issues¶

"Model not found"¶

Problem: Specified model doesn't exist

Solutions:

1. Check model name spelling
2. List available models

python check_api.py --key "$GOOGLE_API_KEY"

Common models:

• veo-3.1-generate-preview
• veo-3.1-fast-generate-preview
• veo-2.0

"Rate limit exceeded"¶

Problem: Too many requests

Solutions:

1. Wait 30-60 seconds between requests
2. Use --preset budget for lower cost
3. Batch requests during off-peak hours

"Billing not enabled"¶

Problem: Video generation requires billing

Solution:

1. Go to Cloud Console
2. Enable billing for your project
3. Link a payment method

Video Extension Issues¶

"video is empty"¶

Problem: External URL used for extension

Cause: API doesn't support HTTP/HTTPS URLs

Solution: Download video first, then use local file

# Download
curl -o local.mp4 "https://example.com/video.mp4"

# Use local file
python video_gen.py "Continue" --extend-video local.mp4

"fps mismatch"¶

Problem: Source video is not 24fps

Solution: Convert to 24fps using ffmpeg

# Check current fps
ffprobe -v error -select_streams v:0 -show_entries stream=r_frame_rate input.mp4

# Convert to 24fps
ffmpeg -i input.mp4 -r 24 -c:v libx264 output_24fps.mp4

# Use converted file
python video_gen.py "Continue" --extend-video output_24fps.mp4

"Model does not support video_extension"¶

Problem: Using veo-3.1 for extension

Solution: Use veo-2.0 for video extension

python video_gen.py "Continue" \
  --extend-video input.mp4 \
  --model veo-2.0

Reference Image Issues¶

"Too many references"¶

Problem: Exceeded reference image limits

Limits:

• Maximum 3 asset images
• Maximum 1 style image
• Cannot mix asset and style

Solution: Reduce number of references

# Use max 3 assets
python video_gen.py "prompt" \
  --reference-image img1.png:asset \
  --reference-image img2.png:asset \
  --reference-image img3.png:asset

"does not support this mix"¶

Problem: Mixing asset and style references

Solution: Use only one type per request

# ❌ Wrong
--reference-image face.png:asset --reference-image style.jpg:style

# ✅ Correct - assets only
--reference-image face1.png:asset --reference-image face2.png:asset

# ✅ Correct - style only
--reference-image painting.jpg:style

"Reference type not supported"¶

Problem: Style reference on wrong model

Solution: Use veo-2.0-exp for style

python video_gen.py "prompt" \
  --reference-image style.jpg:style \
  --model veo-2.0-exp

GCS Issues¶

"Bucket not found"¶

Problem: GCS bucket doesn't exist

Solution:

# Create bucket
gsutil mb gs://your-bucket-name

"Permission denied on bucket"¶

Problem: No access to GCS bucket

Solution:

# Grant access
gsutil iam ch user:your@email.com:objectAdmin gs://your-bucket

"gsutil not found"¶

Problem: gcloud SDK not installed

Solution:

# Install gcloud
curl https://sdk.cloud.google.com | bash

# Or use Python package
pip install google-cloud-storage

General Issues¶

"Timeout"¶

Problem: Request took too long

Solutions:

1. Video generation can take 60-180 seconds
2. Check internet connection
3. Try again - may be temporary

"Output file not created"¶

Problem: Video generation failed silently

Solutions:

1. Check --verbose output for errors
2. Verify output directory exists
3. Check disk space

python video_gen.py "prompt" --verbose

"Import error: requests"¶

Problem: Missing dependencies

Solution:

pip install requests google-auth google-auth-oauthlib

Getting Help¶

1. Check this troubleshooting guide
2. Run with --verbose for details
3. Check Changelog for known issues
4. Open issue on GitHub

Learn More¶

• Authentication - Auth setup guide
• Video Extension - Extension requirements
• GCS Storage - GCS configuration