ElevenLabs v4 Text to Speech

Recorded on 2026-10-02. Four direct requests in this tool's request format (Eleven v4 with Aria and George, v4 Turbo with Jessica, one in Chinese): scripts of 36 to 126 characters returned 6.4 to 9.4 seconds of 44.1 kHz MP3, 3 to 5 seconds after each request. A listener model that was not shown the scripts transcribed all four word for word and described the tagged line as enthusiastic, then whispered. Eight voice-overs made by clicking through this page on a local build were each charged once (2 credits), could be played 8 to 12 seconds after the click, downloaded as MP3 for their owner and returned 401 to a signed-out request. Each request goes to the provider's queue once and is checked on every second, so a slow one does not hold up other jobs. The same request sent twice was charged once, a mismatched price was refused, and a refused, failed, unusable or overdue provider result returned the credits. In a separate check all 20 voices returned audio, and a made-up voice name was refused. Prices follow fal's list price (USD 0.08 per 1,000 characters, 0.04 on Turbo) at 250 credits per USD. Long scripts (direct requests, same day): 3,000 English characters returned about 3 minutes 10 seconds of speech after 47 to 48 seconds on both models; a numbered list of 2,876 Chinese characters returned about 8 minutes 13 seconds after 169 seconds (Turbo) and 219 seconds (v4), and the last 25 seconds matched the end of the script. Through this page, an original Chinese story of 1,018 characters on v4 Turbo cost 11 credits and came back after 40 seconds as 3 minutes 32 seconds of speech whose start, middle and end matched the script. One request is limited to 3,000 characters, with Chinese, Japanese and Korean characters counted as three (1,000 per request), so that a script is three to four minutes of speech and well inside the provider's five-minute limit. Voice cloning, multi-speaker dialogue and longer scripts are not offered here.

• Eleven v4 text to speech on fal — API specification
• Eleven v4 Turbo text to speech on fal — API specification

Updated Oct 2, 2026 · How we test