Skip to main content
POST
/
generation
/
qwen
/
qwen-3-tts
/
text-to-speech
/
1.7b
Qwen 3 TTS 1.7B
curl --request POST \
  --url https://open.skills.video/api/v1/generation/qwen/qwen-3-tts/text-to-speech/1.7b \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "text": "<string>"
}
'
{
  "id": "gen_123",
  "status": "IN_QUEUE",
  "input": {
    "text": "Hello world.",
    "prompt": "Hello world.",
    "voice": "Vivian"
  },
  "usage": {
    "total": 20,
    "subscription": 20,
    "permanent": 0
  }
}
Text-to-speech with preset voices or cloned speaker embeddings and controllable decoding parameters.

Authorizations

​
Authorization
string
header
required

API Key for API endpoints

Body

​
text
string
required

The text to be converted to speech.

Example:

"I am solving the equation x = (-b +/- sqrt(b^2-4ac)) / 2a. Nobody can, it is a disaster and very sad."

​
prompt
string

Optional prompt to guide the style of generated speech. Ignored when speaker embedding is provided.

Example:

"Very happy."

​
voice
enum<string>

Preset voice for synthesis. Ignored when speaker embedding is provided.

Available options:
Vivian,
Serena,
Uncle_Fu,
Dylan,
Eric,
Ryan,
Aiden,
Ono_Anna,
Sohee
Example:

"Vivian"

​
language
enum<string>
default:Auto

The language of the voice.

Available options:
Auto,
English,
Chinese,
Spanish,
French,
German,
Italian,
Japanese,
Korean,
Portuguese,
Russian
Example:

"English"

​
speaker_voice_embedding_file_url
string<uri>

URL to a speaker embedding safetensors file from fal-ai/qwen-3-tts/clone-voice. If provided, cloned voice is used instead of predefined voices.

Example:

"https://storage.googleapis.com/falserverless/example_outputs/qwen3-tts/clone_out.safetensors"

​
reference_text
string

Optional reference text used when creating the speaker embedding. This can improve quality when using cloned voice.

Example:

"Okay. Yeah. I resent you. I love you. I respect you. But you know what? You blew it!"

​
top_k
integer
default:50

Top-k sampling parameter.

Required range: x >= 0
​
top_p
number
default:1

Top-p sampling parameter.

Required range: 0 <= x <= 1
​
temperature
number
default:0.9

Sampling temperature; higher means more random output.

Required range: 0 <= x <= 1
​
repetition_penalty
number
default:1.05

Penalty to reduce repeated tokens/codes.

Required range: x >= 0
​
subtalker_dosample
boolean
default:true

Sampling switch for the sub-talker.

​
subtalker_top_k
integer
default:50

Top-k sampling for the sub-talker.

Required range: x >= 0
​
subtalker_top_p
number
default:1

Top-p for sub-talker sampling.

Required range: 0 <= x <= 1
​
subtalker_temperature
number
default:0.9

Temperature for sub-talker sampling.

Required range: 0 <= x <= 1
​
max_new_tokens
integer
default:200

Maximum number of new codec tokens to generate.

Required range: 1 <= x <= 8192

Response

Generation status

​
id
string
required

Generation id.

​
status
enum<string>
required

Generation status

Available options:
starting,
processing,
succeeded,
failed,
canceled
​
usage
object
required

Credit usage breakdown for the request

​
input
Qwen 3 TTS 1.7B Input · object
required