Voice creation
For fish audio voice cloning, start with a recording you have permission to use and a short script for testing. This guide separates preparing a voice sample from generating speech, so you can check the result before using it in a project.
Have a permitted recording, a quiet place to make a replacement if needed, and a separate test script. Decide whether you need a new voice at all before preparing an upload.
Treat the first attempt as a test, not a finished voiceover. Keep the source and test script separate so you can tell whether a problem comes from the recording or the generated reading.
Record speech from the person who has agreed to its use. Listen for room echo, background music, overlapping speakers, clipping and long silences; make a cleaner recording if these distract from the voice.
Trim unrelated conversation and keep a copy of the original. Before uploading, check the service's current file requirements and consent terms rather than assuming a particular format, duration or upload limit.
Follow the available voice-creation flow, then enter a short, unfamiliar script. Include a question, a name and a longer sentence to hear how the voice handles phrasing beyond the source recording.
Compare the test with the permitted recording for pronunciation, pacing and unwanted noise. If the result misses the mark, improve the source or rewrite the test sentence before making more project audio.
Voice cloning is useful when a particular, authorized speaker matters. An existing voice is often the simpler choice when the project needs readable narration but not that speaker's identity.
Starting material
Create a voice from a permitted sample
A recording of the speaker you are authorized to use
Choose an existing voice
Written text and a voice selection
Best fit
Create a voice from a permitted sample
A consistent, recognizable voice for an approved project
Choose an existing voice
Narration without a specific person's likeness
Permission check
Create a voice from a permitted sample
Confirm the speaker agrees to this use
Choose an existing voice
Review the terms attached to the selected voice
Preparation
Create a voice from a permitted sample
Review and clean the source recording
Choose an existing voice
Compare available voice samples
First test
Create a voice from a permitted sample
Generate a short script not present in the recording
Choose an existing voice
Read the same script with several voices
Likely revision
Create a voice from a permitted sample
Improve the recording or adjust the script
Choose an existing voice
Change the voice or adjust the script
Main limitation
Create a voice from a permitted sample
A poor source can affect the resulting voice
Choose an existing voice
The voice will not reproduce a particular speaker
A voice can sound convincing on one sentence yet struggle across a full script. Use these project checks to catch a mismatch before committing to longer audio.
A source clip includes music beneath every spoken line.
Record clean speech rather than expecting the voice-making process to remove the music. If matching a particular speaker is unnecessary, try a text-first narration workflow.
fish audio text to speechThe team wants a familiar sound but has no permission from the speaker.
Do not use the recording as a source. Select an authorized alternative and test whether it suits the show's tone.
fish audio voice libraryA test line sounds good, but names and emotional dialogue sound uneven.
Test representative lines before producing a full scene. If the project starts with performed dialogue, compare a speech-to-speech approach.
fish audio voice changerA small test makes it easier to hear whether the source recording is suitable. Review the destination's current voice options and requirements before uploading audio; availability and controls may differ from the workflow described here.
Explore voice toolsBegin with a recording you own or have explicit permission to use, then check Fish Audio's current voice-creation instructions and upload requirements. Create a short test from new text and listen for pronunciation, pacing and recording artifacts before using the voice in a longer project.
A voice based on a particular speaker needs suitable source speech and permission to use it. If you only need spoken narration, choosing an existing voice may avoid recording a new sample.
Listen to the source first: music, echo, clipping or another speaker may be audible there. Try a cleaner permitted recording, then test again with the same short script so you can compare the change.
Do not assume that a clip being public gives you permission to reproduce the speaker's voice. Get clear authorization for the intended use and review the service's current rules before uploading it.