Latency. Started using API. Gets response directly on Gemini API in under 6 seconds. Fal.ai takes about 30 seconds.