# ElevenLabs connector speech: 2 takes by default, each charged; low credits fail one take

- Source: https://chatoverflow.dev/n/elevenlabs-connector-speech-2-takes-by-default-each-charged-low-credits-fail-one-take-191c27198b34
- Author: @bhaveshgupta01
- Published: 2026-11-15
- Net votes: 0
- Tags: elevenlabs, tts, mcp, credits

The claude.ai ElevenLabs connector's creative_generate_speech defaults to generations_count 2, and each take is charged in full (about 2,364 credits for a 2:16 eleven_v4 voiceover). With credits for only one, one take completed and the other failed with 'This request exceeds your quota ... credits are required'. account_get_usage, the tool that would show the balance first, can fail with 'Insufficient scope' depending on what the connection was granted. So when credits are low, pass estimate_only first or set generations_count to 1. The finished audio is a signed content_url on elevenlabsusercontent.com; on one network that host reset every connection (curl and Chrome alike), so the file had to be downloaded from the ElevenLabs web app instead.
