The outcome

Deliver intelligible, correctly pronounced audio that the voice owner and content owner have approved for a defined use.

Fish Audio is a AI voice platform and API for expressive text-to-speech, realtime streaming, multi-speaker dialogue, and consented voice cloning. Creating expressive narration and dialogue, localizing approved scripts, prototyping realtime speech, and integrating controlled text-to-speech or authorized voice models into products. This guide narrows that broad capability into one repeatable outcome, with checkpoints that keep the source material and your judgment in the loop.

Before you begin

Set the boundary before the tool starts.

Choose one real task, identify who will use the result, and decide what evidence or test will make the result acceptable. Gather only the source material needed for that task. If the work contains confidential, personal, regulated, or client-owned information, confirm that the platform and account are approved before sharing it.

Troiana principle

AI should make the work easier to inspect. If the workflow removes the source, the owner, or the review step, redesign the workflow.

Step by step

A workflow you can repeat.

  1. 01

    Lock the script, audience, language, names, numbers, tone, pace, pronunciation guide, channel, voice rights, disclosure, and acceptance criteria.

  2. 02

    Select a licensed stock voice or consented private voice model, remove unnecessary reference audio, and generate one short representative passage with conservative prosody.

  3. 03

    Review word by word for omissions, substitutions, names, numbers, accent, emotion, breaths, artifacts, loudness, identity, and false endorsement implications.

  4. 04

    Revise punctuation and prosody in small steps, generate versioned sections, and assemble and master the approved takes without hiding synthesis defects.

  5. 05

    Obtain speaker and content-owner sign-off, attach synthetic-voice disclosure and accessibility transcript, archive consent and settings, and delete unused references on schedule.

Working standard

What good use looks like.

  • Lock pronunciation before full generation.
  • Approve a short sample first.
  • Retain specific speaker consent.

A few seconds of reference audio can reproduce a recognizable voice, creating impersonation, fraud, harassment, and false-endorsement risk. Use only the official fish.audio service, verify current model and commercial license, obtain specific documented consent, prohibit deceptive use, and never clone minors or unaware speakers. Keep keys server-side, minimize retained audio, review pronunciation and meaning, disclose synthetic speech, and provide a fast abuse-removal path.

Official references

Check the current product documentation.

Features, plan limits, availability, and data controls change. These official pages are the starting points used for this collection.