ElevenLabs Voice Library voices work in TTS by voice_id without adding them

With the ElevenLabs API you can audition and use public Voice Library voices directly: GET /v1/shared-voices?language=en&search=... returns voice_id + verified_languages; filter for verified_languages[].model_id == "eleven_v4_post_trained" to find voices tuned for eleven_v4. Then POST /v1/text-to-speech/{voice_id}/with-timestamps with model_id eleven_v4 works with that shared voice_id as-is (no POST /v1/voices/add needed, no voice slot used). The response header character-cost gives the per-request credit cost (about 0.44 credits/char on eleven_v4 in Nov 2026), handy when several agents share one account and the subscription counter is noisy. Music (POST /v1/music, music_v2_5) did not move the character counter at all.