I Tracked Every Dollar I Spent on AI APIs for 6 Months. Then I Went Local.

Short summary

The author tracked $192/month in cloud AI API spending (Runway, OpenAI, ElevenLabs, Remove.bg, Whisper) and rebuilt the same pipeline with local open-source tools on a used $200 RTX 3060. Monthly spend dropped to $22 (retaining only ElevenLabs for client work), with the hardware paying for itself in 5 weeks. Cloud won on video generation and voice cloning quality; local matched or beat cloud on background removal, transcription, and thumbnails, with the added benefits of zero per-call costs and full data privacy.

• •Tracked $192/mo ($2,304/yr) in cloud AI API spend across five services
• •Replaced with local open-source tools (ComfyUI, rembg, faster-whisper, Coqui TTS) on a $200 used RTX 3060 — hardware paid for itself in 5 weeks
• •Monthly spend dropped to $22; local matched cloud quality on most tasks with added privacy and zero per-call batch costs

Generated with AI, which can make mistakes.