Skip to main content
POST
Fish Audio Voice Cloning
Fish Audio API for creating a voice model (voice cloning).

Request Headers

string
required
Enum: application/json
string
required
Bearer authentication format, for example: Bearer {{API Key}}.

Request Body

enum<string>
required
Model type, tts is for text to speech.Available options: ttsAllowed value: "tts"
string
required
Model title or name.
enum<string>
required
Model train mode, for TTS model, fast means model instantly available after creation.Available options: fastAllowed value: "fast"
file[]
required
Upload voices files that will be used to tune the model.
enum<string>
default:"public"
Model visibility, public will be shown in the discovery page, unlist allows anyone with the link to access, private only be visible to the creator.Available options: public, unlist, private
string | null
Model description.
file | null
Model cover image, this is required if the model is public.
string[]
Texts corresponding to the voices, if unspecified, ASR will be performed on the voices.
string[]
Model tags.
boolean
default:false
Enhance audio quality.

Response

string
required
Unique identifier for the created model.
enum<string>
required
Model type.Available options: svc, tts
string
required
Model title or name.
string
required
Model description.
string
required
URL of the model cover image.
enum<string>
required
Current state of the model.Available options: created, training, trained, failed
string[]
required
Model tags.
string<date-time>
required
Timestamp when the model was created.
string<date-time>
required
Timestamp when the model was last updated.
enum<string>
required
Model visibility setting.Available options: public, unlist, private
integer
required
Number of likes the model has received.
integer
required
Number of marks/bookmarks the model has received.
integer
required
Number of times the model has been shared.
integer
required
Number of tasks associated with the model.
AuthorEntity · object
required
Information about the model author.
enum<string>
default:"full"
Training mode used for the model.Available options: fast, full
SampleEntity · object[]
Sample data associated with the model.
string[]
Languages supported by the model.
boolean
default:false
Whether the visibility setting is locked.
boolean
default:false
Whether the current user has unliked the model.
boolean
default:false
Whether the current user has liked the model.
boolean
default:false
Whether the current user has marked/bookmarked the model.
Last modified on May 6, 2026