Skip to main content
SUNO allows us to create custom voice personas through any audio file, enabling voice cloning for music generation. Unlike the existing Persona API (which uses Suno-generated audio_id), this API accepts a publicly accessible audio_url, namely your own vocal recording. This document explains how to integrate the voice cloning API.

​
Step 1: Create a Voice Persona

This API has three input parameters: audio_url (required), which is a publicly accessible MP3 or WAV audio file URL containing clear vocals from a single person; name and description (optional), which are the name and description of the voice persona.
Audio File Requirements
  • The audio format must be WAV or MP3
  • The audio duration must be between 10~240 seconds; clean single-speaker dry vocal material of 30~60 seconds is recommended
  • The audio should contain clear, recognizable speech or singing vocals from a single person
  • Be sure to avoid background noise, accompaniment, echo, and reverb; complete songs with accompaniment will usually fail voiceprint verification
  • Do not include multiple speakers or multiple vocal layers
  • Material with volume that is too low, unclear speech, or excessive noise may result in cloning failure or poor generation results Usage Restrictions
  • Voice personas created by uploading audio are private resources
  • This voice persona does not support reuse across accounts
  • It is recommended to use it as soon as possible after successful creation; it may become invalid or unavailable if left unused for a long time
  • The returned name is automatically generated by the system; please use the returned persona_id as the reference
Please Retry First When a Call Fails Voice cloning is a compute-intensive task. Even if the material is fully compliant, there is still a certain probability of occasional failures, with common responses such as voices_sound_different (voiceprint verification failed). Such failures are unrelated to audio quality, and retrying with the same material can usually succeed. It is recommended to implement 1~2 automatic retries for failed results during integration. Failed requests will not be charged.
The above https://cdn.acedata.cloud/suno_demo.mp3 is an example material that can be called directly (MP3, 41 seconds, single-person dry vocal). If you need a WAV format example, you can use https://cdn.acedata.cloud/uploads/82d23b97-ec1c-4b41-91b8-989fc51f8765 (WAV, 41 seconds, mono 44.1kHz).
The result is as follows:
As you can see, the persona_id field in data is the created voice persona ID. The is_public field is always false, because voice personas created by uploading audio are private. Note that the returned name is automatically generated by the system; please use persona_id to reference this voice persona subsequently.

​
Step 2: Generate Music Using the Voice Persona

After obtaining the voice persona ID, we can use the Suno Audios Generation API for music generation. Set action to generate, and set persona_id to the voice persona ID returned above; the generated song will be sung using the cloned voice.
Note: Voice cloning only supports chirp-v4-5 and later models (such as chirp-v4-5, chirp-v5, and chirp-v5-5), and does not support chirp-v4.
The result is as follows:
persona_id can be used for generate generation and also for cover / upload_cover covers. There is no need to specify vocal_gender at the same time when specifying a voice; voice similarity still depends on the reference material and generation result. Voices created through this interface are private resources. Before covering an existing song, first call /suno/upload, passing the original song’s audio_url and the voice’s persona_id; after a successful upload, call /suno/audios using the returned data.audio_id and the same persona_id:
Creating a voice, uploading the original song, and creating the cover must use the same service channel. Do not directly reuse a song ID that was not previously associated with this private voice; see Suno Upload Reference Audio for details.