Disclosure: This article was written with AI assistance and reviewed by EmpirioLabs AI.
Short answer: for the same 164-character narration, Qwen Audio 3.1 TTS cost the least, about a twentieth of a cent, StepAudio 3 TTS returned its file fastest, and Gemini 2.5 Pro TTS cost the most, about four cents, with the slowest and longest read. Qwen Audio 3.1 TTS Plus, the quality tier of Qwen Audio 3.1, cost just under one cent. All four run on EmpirioLabs through the same speech endpoint. Listen to each clip below; we are not declaring a winner.
Listen to the four voices
The narration: "The lighthouse keeper wrote one line in the log every night for thirty-one years. On the last page, in a different hand, someone had written: the light is still on." The direction: "A documentary narrator: calm and unhurried, then a quiet, moved pause before the final sentence."
Results at a glance
| Qwen Audio 3.1 TTS Plus | Qwen Audio 3.1 TTS | Gemini 2.5 Pro TTS | StepAudio 3 TTS | |
|---|---|---|---|---|
| Maker | Alibaba | Alibaba | StepFun | |
| Billing unit | Characters | Tokens: text in, audio out | Tokens | Characters |
| Price for this request | $0.0098 | $0.0005 | $0.0433 | $0.0024 |
| Audio length | 11.2 seconds | 10.0 seconds | 24.0 seconds | 9.7 seconds |
| Time to the finished file | 14.3 seconds | 12.8 seconds | 16.0 seconds | 3.7 seconds |
| Voices | 4 | 19 | 30 | 12 |
| Direction field | instruction | instruction | style_prompt | instruction |
How we ran it
Each model read the identical narration with the identical direction, passed in the model's own direction field. Each used an English-speaking male voice from its own library and returned MP3 (Gemini 2.5 Pro TTS returns WAV, converted to MP3 for this page), with every other setting at its default. One shot each. Prices are the amounts charged for these exact requests. Times are wall-clock seconds for one request on November 2, 2026, from sending the request to receiving the finished file; all four models also stream audio while it is generated.
What to look for
The pause. The direction asks for a pause before "the light is still on". Listen for where each voice places it and how long it holds; Gemini 2.5 Pro TTS held a pause of about a second and a half, and its whole read ran more than twice as long as the others.
Plus against Flash. Qwen Audio 3.1 TTS Plus and Qwen Audio 3.1 TTS take the same request body. Plus is the tier Alibaba builds for narration, dubbing and brand voices; listen to the two Qwen clips side by side for the difference in delivery.
Length and price. Qwen Audio 3.1 TTS and Gemini 2.5 Pro TTS bill the audio they generate, so a slower read costs more; TTS Plus and StepAudio 3 TTS bill the characters you send.
Run the same test on EmpirioLabs
curl https://api.empiriolabs.ai/v1/audio/speech \
-H "Authorization: Bearer $EMPIRIOLABS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen-audio-3-1-tts-plus",
"voice": "xunanchuan_v3.1",
"input": "The lighthouse keeper wrote one line in the log every night for thirty-one years. On the last page, in a different hand, someone had written: the light is still on.",
"instruction": "A documentary narrator: calm and unhurried, then a quiet, moved pause before the final sentence.",
"response_format": "mp3"
}'Swap model and voice to run the same request against qwen-audio-3-1-tts, gemini-2-5-pro-tts (with style_prompt in place of instruction) or stepaudio-3-tts, or try them in the playground.
Frequently asked questions
Were the clips edited or retried?
No. Each clip is the first audio the model returned for this request, unedited.
Why does Qwen Audio 3.1 TTS Plus cost more than Qwen Audio 3.1 TTS?
They are priced differently: Plus per character of text, the Flash tier by tokens. For this narration, Plus cost about twenty times as much.
Which one should I use?
Listen to the clips with your own script. The same request works for all four with only the model, voice and direction field changed. The launch guide, How to Use the Qwen Audio 3.1 TTS Plus API, covers its voices and custom voices.
