Help

Voice Studio

Create and manage the voices that narrate your stories — clone a voice, design one from a description, or pick one from the ElevenLabs library.

Updated 23 October 2026

What it does

Voice Studio is where your voices live. You make a voice once here, and then use it in any story: the story's Voice step only picks a voice and generates the audio. On the left is My Voices, on the right the panel for creating a voice or building effects.

Creating a voice is free. Credits are charged when a voice speaks a script — per character for ElevenLabs voices, per minute of audio for the other engines. The price shows on the generate button before you click.

Create a voice

Click + above My Voices and choose how:

  • ElevenLabs (recommended) — usually the best quality. Inside you choose again:
    • Clone from audio — upload one or more recordings of the voice.
    • Voice library — search ElevenLabs' public voices by language, accent, category, gender and quality, listen to them, and add one.
    • Design a voice — describe the voice in words and pick one of the previews. Currently unavailable: our ElevenLabs account is at its voice limit.
  • Built-in voices — ready-made voices that run on our own server: fast, no waiting. Pick one, or blend a few together with a mix ratio for each.
  • Clone from audio — clones a voice from your recording on a cloud GPU. High quality (cloud) needs the exact transcript of the recording and comes very close to the original. The beta engine needs no transcript but sounds less like the speaker.
  • Design a voice — describe the voice in up to 500 characters, e.g. "Calm middle-aged male narrator, warm and slightly husky, moderate pace". Good descriptions can be saved as presets.

Every voice gets a name, an optional description and its language.

How long generating takes

  • ElevenLabs and built-in voices: much faster than the audio itself.
  • Cloned voices on the cloud GPU: about a third to half of the audio's length (10 minutes of audio ≈ 3–5 minutes).
  • Designed voices: about as long as the audio itself.
  • The first generation after a pause adds a one-time cold start.

Manage your voices

Each voice in My Voices shows its type and engine. The buttons on a voice:

  • ▶ Preview — hear a short sample.
  • Tag — group voices, e.g. all voices of one movie.
  • Edit — name, description and language. For ElevenLabs voices also the model and how it speaks: Speed, Stability, Similarity, Style and Speaker boost. Changes are only kept when you click Save.
  • Make public — every Seedflows user can then find and use the voice. Make private takes it back.
  • Export — downloads the voice as a .tts.zip file. Import (the upload button above the list) brings it back, e.g. into another account.
  • Delete — removes the voice.

Search and the model filter above the list help when you have many voices.

Tips for ElevenLabs settings

  • Lower Stability = more emotion and variety; higher = calmer and more even.
  • Higher Similarity = closer to the original recording, but also to its noise.
  • Style exaggerates the voice's character — keep it low for long narration.

Effect presets

The tune button above the list opens Effect presets: reusable sound effects that work on any voice, e.g. Radio, Telephone, Megaphone, Robotic, Underwater, Echo Chamber or a Deep voice. Start from a built-in or from scratch with New effect, adjust reverb, echo, pitch, filters and compression with sliders, and Test it against one of your voices. Built-in presets can't be changed — saving one makes your own copy.

In a story's Voice step you then pick an effect for a speaker, e.g. for a character that speaks over the phone.

Good recordings for cloning

  • 30 seconds to a few minutes of one speaker, no music, no other voices.
  • A quiet room, no echo, steady distance to the microphone.
  • Speak the way the voice should sound later — a calm reading clones as a calm voice.
  • For High quality (cloud): type exactly what is said in the recording, word for word.

When something goes wrong

  • The clone doesn't sound like the speaker — use a cleaner, longer recording, check the transcript, or clone it with ElevenLabs.
  • Odd sounds or artefacts in a blend of two voices — the mix slider goes from −100 % to 200 %; below −50 % or above 150 % it can cause artefacts. Bring it back toward the middle.
  • A voice is missing in a story — in dialogues all speakers must use the same engine; the Voice step then only lists matching voices.
Open Voice Studio Still stuck? Contact us

More help

Face StudioKeep the faces of your recurring characters in one place — upload a photo or generate a face — so every scene and avatar shows the same person.Avatar StudioCreate a consistent on-screen character — dress it, place it in a scene — and make it talk from text, a story or your own audio.Image StudioGenerate a single image outside a story — with your style, rules and up to two people or props as references so they stay recognisable.Video StudioMake a single AI video clip outside a story — from a description, optionally starting from your own picture or moving from a start to an end picture.Motion StudioDescribe a motion graphic — or let a voice over drive it — and an AI agent designs, animates and renders it as an MP4.Product StudioTurn what a channel's viewers want into an ebook, sell it on your own Gumroad account and on a ready-made sales page for your video descriptions.Uploading to YouTubePrepare title, description, tags and thumbnail for a story and upload the finished video to your YouTube channel.