Create speech
OpenAI-compatible text-to-speech (MiniMax Speech 2.8 HD / Turbo, Doubao Speech); the response is the audio bytes directly.
/v1/audio/speechSynthesizes text into audio synchronously; the response is the audio itself, with no task to poll. Works directly with the OpenAI SDK's audio.speech.create; only the Base URL and model / voice change. The two speech models belong to the separate MiniMax Speech Official group; a key must include that group to call them, and you can add it to an existing chat or video key.
Any other field returns 400 unknown field, including stream_format, sample_rate, language_boost and voice_setting. For those, streaming output or cloned voices, use the native endpoint POST /v1/t2a_v2; see the Text-to-speech guide. Billed per character: one Han character counts as 2 characters.
Doubao Speech seed-tts-2.0 (text to speech) and seed-audio-1.0 (audio from a scene description) belong to the separate Doubao Speech Cloud group and use only this endpoint: input is counted in characters, up to 5,000 for seed-tts-2.0 and 3,000 for seed-audio-1.0; speed is 0.5–2.0; the two models accept different response_format values; and failed requests are not billed. Switch the model picker above to "Doubao Speech" for its parameters; see the Text-to-speech guide · Doubao Speech.
Doubao Voice Cloning seed-icl-2.0, in the same group, also uses this endpoint: first clone a voice with Clone voice to get a voice_id starting with hbv_, then pass it as voice. Parameters and limits are the same as seed-tts-2.0; it is billed by input characters at the official price of ¥3 per 10,000 characters. See Doubao voice cloning.
The body is audio bytes, so the actual charge for this request is returned only in the X-HopBase-Cost / X-HopBase-Cost-USD / X-HopBase-Currency headers. See Account API · Charges.
Headers
Bearer sk-…: an API key created in the console under API keys; its group must include the requested model
Body parametersJSON
Case-sensitive, no aliases; tts-1 / tts-1-hd are not recognized
Valuesspeech-2.8-hdspeech-2.8-turboseed-tts-2.0seed-audio-1.0seed-icl-2.0
Text to synthesize (the native endpoint /v1/t2a_v2 allows up to 10000 characters)
Length1–4096 chars
A system voice ID, a cloned voice ID, or an OpenAI voice name: alloy, ash, ballad, cedar, coral, echo, fable, marin, nova, onyx, sage, shimmer, verse (OpenAI voice names map to the default voice). Default: Chinese (Mandarin)_News_Anchor if the text contains Han characters, otherwise English_expressive_narrator
aac returns 400; opus is 24 kHz in an Ogg container
Valuesmp3opusflacwavpcm
Default"mp3"
Out-of-range values return 400; in-range values are clamped to 0.5–2
Range0.25–4Default1
Accepted but ignored
Case-sensitive, no aliases; tts-1 / tts-1-hd are not recognized
Valuesspeech-2.8-hdspeech-2.8-turbo
Text to synthesize (the native endpoint /v1/t2a_v2 allows up to 10000 characters)
Length1–4096 chars
A system voice ID, a cloned voice ID, or an OpenAI voice name: alloy, ash, ballad, cedar, coral, echo, fable, marin, nova, onyx, sage, shimmer, verse (OpenAI voice names map to the default voice). Default: Chinese (Mandarin)_News_Anchor if the text contains Han characters, otherwise English_expressive_narrator
aac returns 400; opus is 24 kHz in an Ogg container
Valuesmp3opusflacwavpcm
Default"mp3"
Out-of-range values return 400; in-range values are clamped to 0.5–2
Range0.25–4Default1
Accepted but ignored
Case-sensitive, no aliases; tts-1 / tts-1-hd are not recognized
Valuesseed-audio-1.0seed-icl-2.0seed-tts-2.0
Counted in characters (a Chinese character counts as 1): at most 5,000 for seed-tts-2.0 and seed-icl-2.0, and 3,000 for seed-audio-1.0; for seed-audio-1.0, write "scene description + lines"
Length1–5000 chars
A Doubao voice ID such as zh_female_vv_uranus_bigtts, zh_male_m191_uranus_bigtts, en_female_dacey_uranus_bigtts. For seed-tts-2.0, omitting it or sending an OpenAI voice name (such as alloy) picks the default by text language: zh_female_vv_uranus_bigtts for Chinese, en_female_dacey_uranus_bigtts otherwise; for seed-audio-1.0, omitting it or sending an OpenAI voice name lets the description decide the voice. An unknown voice returns 400 unknown_voice. seed-icl-2.0 requires the cloned voice ID (starting with hbv_) returned by POST /v1/voice_clone; omitting it or sending another voice returns 400 voice_required, and a cloned voice used with any other model returns 400 unknown_voice
seed-tts-2.0 and seed-icl-2.0: mp3 / opus / aac / wav / pcm; seed-audio-1.0: mp3 / wav / pcm / flac; anything else returns 400
Valuesmp3opusaacwavpcmflac
Default"mp3"
0.5–2.0; values out of range return 400
Range0.5–2Default1
Accepted but ignored
Returns
200Audio bytes. Content-Type follows response_format: audio/mpeg, audio/ogg, audio/flac, audio/wav, audio/pcm
Response headers
The charge for this request in CNY (the same charge as X-HopBase-Cost-USD), up to 6 decimal places, such as 0.0816. Sent only when the request is charged; not sent on streaming (SSE) responses. When image generation takes longer than about 40 seconds the headers have already been sent, so read usage.cost in the body instead
The USD amount actually deducted from your balance for this request, the same amount as its usage record, up to 8 decimal places, such as 0.012; sent only together with X-HopBase-Cost
The currency of X-HopBase-Cost, always CNY
audio/mpeg Audio bytes
Errors
unknown field, input too long, speed out of range, unsupported response_format; for Doubao Speech, an unknown voice returns unknown_voice. For seed-icl-2.0: no cloned voice returns voice_required, a voice still being created or trained returns voice_not_ready, and a voice whose training failed returns voice_training_failedmissing_api_key / invalid_api_key / api_key_expired)insufficient_quota)model_not_found), or the path does not belong to the group (route_not_found)request_too_large)user_concurrency_limit / apikey_concurrency_limit), with Retry-After