Text-to-speech model pricing compared
Every voice BibiGPT can narrate with, priced per million characters, with an official sample to listen to. Token-billed models are converted at about 15 characters per second of speech.
Prices checked 2026-10-26. These are upstream list prices, not BibiGPT pricing.
| Model | Listen | per 1M chars | 1M chars | List price | Strengths | In BibiGPT |
|---|---|---|---|---|---|---|
Kokoro-82M (in-browser) Open source · on device kokoro-jsFree; text stays in the browser. Unsupported languages fall back to system speech. | $0 | $0 | Free (runs in your browser, one-time ~86MB download) | zh/en/ja Low latency | On device In-browser read-aloud | |
Qwen-Audio-3.1 TTS FlashCheapest Alibaba Qwen qwen-audio-3.1-tts-flash3.1 moved to token billing; BibiGPT still runs 3.0 Flash. | $2.87Est. | $2.87 | ¥1.5 input / ¥12 output per 1M tokens (Beijing region) | zh/en + dialect voices Voice design | Not in BibiGPT | |
xAI TTS xAI (hosted on fal) xai/tts/v1The cheapest per-character option we use. | $15 | $15 | $0.015 per 1K characters | Multilingual, Chinese on request Low latency | In BibiGPT | |
Qwen-Audio-3.0 TTS Flash Alibaba Qwen (OpenRouter) qwen/qwen-audio-3.0-tts-flashCheap, steady Chinese narration. | $15 | $15 | $15 per 1M characters (OpenRouter); China site ¥1 per 10K characters, one Chinese character counts as 2 | zh/en + dialect voices | In BibiGPT | |
Fish Audio S2.1 Pro Fish Audio fish-audio/s2.1-proBilled per byte: $15 per million English characters, about $45 per million Chinese characters. Solid voice cloning. | $15 | $15 | $15 per 1M UTF-8 bytes (a Chinese character is about 3 bytes) | 80+ languages Voice cloneExpressive tags | In BibiGPT | |
HeyGen Voice HeyGen (OpenRouter) heygen/heygen-voice-1Reads Chinese and English; default voice is Nadine. | $15 | $15 | $15 per 1M characters (OpenRouter; turbo tier $4) | Multilingual | In BibiGPT | |
Gemini 3.8 Flash TTS Google gemini-3.8-flash-ttsDefault narration voice. Multi-speaker in one call; steer tone with plain-language prompts. | $15.1Est. | $15.1 | $0.50 text input / $9 audio output per 1M tokens (doubling to $1 / $18 from 2027) | 80+ languages Multi-speakerExpressive tags | In BibiGPT | |
GPT-4o mini TTS OpenAI gpt-4o-mini-ttsTone can be steered by instructions; Chinese has a slight accent. | $16.2Est. | $16.2 | $0.60 text input / $12 audio output per 1M tokens (about $0.015/minute) | Multilingual Expressive tags | Not in BibiGPT | |
DeepSeek POST /v1/tools/ttsOnly two voices (Stella, Tide) and no speed or emotion control. About 1 s to first audio and ~5× realtime in our test. Chinese counts double, so about $45 per million Chinese characters. | No official sample | $22.5 | $22.5 | $0.45 per 10K characters (limited-time 50% off: $0.225); a Chinese character counts as 2 | Mandarin, Cantonese, English, Japanese, Korean and 5 more Low latency | In BibiGPT |
Deepgram Aura-2 Deepgram aura-2Low-latency English TTS aimed at voice agents. | $30 | $30 | $0.030 per 1K characters | English and a few others Low latency | Not in BibiGPT | |
Eleven Flash v2.5 / v4 Turbo ElevenLabs eleven_flash_v2_5 · eleven_v4_turbo~75–100 ms to first audio. Pick it from the ElevenLabs model menu when narrating. | No official sample | $40 | $40 | $0.04 per 1K characters | 32 languages (v4 Turbo: 90+) Voice cloneLow latency | In BibiGPT |
Eleven v4 / v3 ElevenLabs eleven_v4 · eleven_v3The most expressive tier; Chinese quality depends on the voice. | $80 | $80 | $0.08 per 1K characters | 70+ languages Voice cloneExpressive tagsMulti-speaker | In BibiGPT | |
MiniMax Speech 2.8 HD MiniMax speech-2.8-hdMost natural Chinese delivery; the Turbo tier is $60 per million characters. | $100 | $100 | $100 per 1M characters (international); China site ¥3.5 per 10K characters, one Chinese character counts as 2 | 40 languages, best in Chinese Voice cloneVoice designExpressive tags | In BibiGPT | |
Seed Audio 1.0 ByteDance Seed (hosted on fal) bytedance/seed-audio-1.0Billed per minute of audio. Describe the voice in a sentence; no reference clip needed. | $208Est. | $208 | $0.1875 per minute of audio (fal); Volcengine China ¥1/minute | Chinese, English Voice design | In BibiGPT | |
MiMo V2 TTS Xiaomi mimo-v2-ttsNatural spoken Chinese. | TBC | — | Not published (the official price page only lists v2.5 TTS as free for a limited time) | Chinese, English Expressive tags | In BibiGPT |
- Gemini / Qwen token-billed TTS: 25 audio tokens per second, about 15 characters per second of speech.
- Chinese text is usually billed per character; some vendors count one Chinese character as 2. Your bill is what counts.
- CNY ↔ USD at 7.1.
FAQ
Which transcription model is cheapest?
On our samples Qwen-Audio-3.1 file transcription came to about ¥0.075–0.10 per hour, below StepAudio 2.5 (¥0.15/h) and Paraformer v2 (¥0.288/h). On-device models in the Bibi Local desktop app cost nothing.
What is the difference between STT, live and TTS?
STT turns a finished recording into text and is billed per audio hour. Live models do the same while you speak, or answer you in voice, and cost more. TTS goes the other way: text in, speech out, usually billed per character.
Why are some prices marked “Est.”?
Those models bill by tokens, not minutes. We convert with the token densities listed under the table; your real bill moves with speech speed and language.
Can I use these models in BibiGPT?
Models marked “In BibiGPT” are covered by membership. “Your key” models are selectable once you paste your own API key. “On device” models run free in the Bibi Local desktop app. “Not in BibiGPT” rows are listed for reference.
Hear these voices on your own content
Turn a summary into a narrated image video in BibiGPT and pick the voice engine there.