Qwen Audio 3.1 TTS Next | EmpirioLabs AI Docs

Alibaba Cloud · Audio Generation

POST /v1/audio/speech

Audio generation from a written description, with narration, dialogue in up to three cloned voices, sound effects, ambience, and music in one file.

At a glance

Field	Value
Model id	`qwen-audio-3-1-tts-next`
Model release date	2026-09-23
Input modalities	Text, Audio
Output modalities	Audio
Context window	-
Weight precision	-
Max output tokens	50,000
Region	China
Features	text_to_speech, audio_generation, multi_speaker, voice_cloning, sound_effects, podcast
Native inference	No
New	Yes
Supported endpoints	`POST /v1/audio/speech`
Alternate model ids	`qwen-audio-3.1-tts-next`, `alibaba/qwen-audio-3-1-tts-next`

Pricing

Charge	Spec	Rate
Input	per 1M prompt tokens	$1.86
Output	per 1M generated tokens	$3.74

Example request

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-next", "input": "Hello from EmpirioLabs."}'

Parameters

Parameter	Type	Required	Default	Description
`input`	string	yes	-	A description of the audio to generate, up to 3,000 characters: the voices, what they say, the delivery, and any sound effects, ambience, or music. Name reference clips as @voice1 to @voice3. · Max: 3000
`references`	array	no	-	Up to 3 voice reference clips as URLs or base64 audio data URIs, each up to 30 seconds and 10 MB, in WAV, MP3, or OGG Opus. Name them @voice1 to @voice3 in input, in order.
`response_format`	enum	no	`"mp3"`	Audio container. mp3 is compressed, wav is uncompressed PCM in a RIFF header, and pcm is headerless raw samples. · Allowed: `mp3`, `wav`, `pcm`
`sample_rate`	enum	no	`48000`	Output sample rate in Hz. 48000 suits music and ambience; 24000 is enough for speech alone. · Allowed: `8000`, `16000`, `24000`, `44100`, `48000`
`channels`	enum	no	`2`	1 for mono, 2 for stereo. · Allowed: `1`, `2`
`speed`	number	no	`1`	Speaking rate multiplier. 0.5 is half speed and 2.0 is double speed. · Range: 0.5 – 2.0
`volume`	integer	no	`50`	Output loudness, where 50 is the reference level. · Range: 0 – 100
`seed`	integer	no	-	Generation seed. Reuse a seed with the same input and settings for a repeatable result. · Range: 0 – 2147483647
`bit_rate`	integer	no	-	Constant MP3 bitrate in bps. Applies to the mp3 format only; without it, MP3 uses a variable bitrate. · Range: 32000 – 320000
`watermark`	boolean	no	false	Add an AI-generated content watermark to the audio.

Notes

What it makes

One finished audio file from a written description: narration, dialogue, sound effects, ambience, and music together.
Describe each voice, the delivery, and the sounds in input. There is no voice list.
Chinese and English.

Voices from reference clips

references takes up to 3 clips as URLs or base64 audio data URIs, each up to 30 seconds and 10 MB, in WAV, MP3, or OGG Opus.
Name them in input as @voice1, @voice2, and @voice3, in order, for example: @voice1 cheerfully says: "What a lovely day."

Input limits

Up to 3,000 characters per request.
instruction and SSML are not used: all direction goes in input.

Output

MP3 (the default), WAV, or PCM at 8,000 to 48,000 Hz, mono or stereo; the defaults are 48,000 Hz and stereo.
Up to 120 seconds per request, or 240 seconds for a podcast.
One complete file per request. This model does not stream.

Billing

Input tokens are the description plus each reference clip. Output tokens are the generated audio, about 200 per second.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/qwen-audio-3-1-tts-next.