Text to Speech API on KeyoAPI — IndexTTS-2, ElevenLabs & Async TTS

Choose the model’s endpoint and complete request fields: Qwen uses prompt/language arrays; CosyVoice and Index require reference audio; Gemini uses native generateContent.

Text to Speech API on KeyoAPI is a model-family pricing lane on KeyoAPI — a prepaid AI API gateway where one Bearer key and one wallet ($1 = 500,000 quota units) cover chat, speech, vision, OCR and video models.

TTS models on Keyo

Indicative figures for planning. Confirm live Keyo sell rates on interactive Model Square or the static pricing list.

Prototype on gift-credit chat, then meter TTS

Gift-credit chat twins (Atria-dawn-v2:free and other :free catalog IDs, billed from your $10 welcome credit) sit on the same key for script drafting. TTS rows above are metered — confirm live rates on Model Square. Guides: the IndexTTS-2 model page, the CosyVoice3 model page, the Qwen3-TTS model page.

Rates and endpoints

TTS models do not share one interchangeable request body. GLM-TTS, Step-Audio-TTS-3B, IndexTTS-2 and Eleven use POST /v1/audio/speech. Qwen3-TTS, CosyVoice3 and Spark-TTS-0.5B use POST /v1/async/audio/speech, then singular GET /v1/task/{id}.

Qwen3-TTS requires an inputs array of objects containing prompt and language (Chinese or English); speaker and instruction are optional model-specific controls. CosyVoice3 requires string inputs plus prompt_wav_url and its exact prompt_text, with a reference ≤30 seconds, ≥16 kHz, WAV/MP3. IndexTTS-2 requires prompt_audio_url and a chosen text/audio emotion branch. See the API endpoint reference for complete examples.

GLM reference examples use prompt_audio_url plus its matching prompt_text; Step references are optional WAV. There is no confirmed interchangeable alloy/echo voice list for these models. Spark basic synthesis explicitly supplies gender, with pitch and speed from one to five. Eleven uses model-specific built-in names such as george; do not copy another service’s voice ID. Request success does not guarantee language, gender, emotion or voice quality.

Gemini 3.8 Flash TTS uses only POST /v1beta/models/gemini-3.8-flash-tts:generateContent, not /v1/audio/speech. Its JSON response contains inlineData.mimeType and base64 data; decode audio according to MIME and wrap raw PCM when needed. See the API endpoint reference for the decoder. The preview TTS model and keyo-tts remain pending; laya is not TTS and InfiniteTalk needs existing audio.

Replace reference URL placeholders with real downloadable files and exact transcripts before submitting. Save task IDs; retry temporary task_query_unavailable GETs according to Retry-After without creating replacement jobs. Check current rates and account groups on the pricing page.

curl https://www.keyoapi.xyz/v1/async/audio/speech \
  -H "Authorization: Bearer $KEYO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"Qwen3-TTS","output_format":"mp3","inputs":[{"prompt":"Hello from KeyoAPI TTS.","language":"English","speaker":"Vivian"}]}'
# Save id/task_id, then GET /v1/task/{id}.

Get started in five steps

1. Create an account — new users receive $10 of gift credit for prototyping.
2. Create an API key in the console.
3. Point your OpenAI SDK at https://www.keyoapi.xyz/v1 with the model ID from this page, or call the dedicated REST path from the API handbook.
4. Track spend in the console Wallet — $1 equals 500,000 quota units, and gift credit burns before recharged balance.
5. Set a per-key spending cap before wiring production traffic.

A verified first call is the acceptance test for the setup: the response should carry the payload shape documented in the handbook, the Wallet ledger should show a deduction in quota units (500,000 units = $1), and repeated calls should deduct linearly at the listed rate. Any mismatch usually means the model ID or endpoint is wrong — check both against the pricing list before debugging client code.

FAQ

Also see the model comparison page, the gift-credit models page, the full pricing list, and CosyVoice3 API · Voice cloning · IndexTTS-2.