Write the script and create the narration.

Design AI voice

Speakeasy creates a voice profile from a written voice description. Use this when a generated voice is acceptable or preferred.

Clone from sample

Upload or record a clean voice sample before writing the final script. Add an optional reference transcript and style instruction for better matching.

Use final audio

Upload or record finished narration. This skips the Script & Voice step because the final audio is already supplied.

1. 01

   Add a prompt

   Describe what the presenter should say. Add an optional target runtime in seconds when the script needs to fit a specific length.
2. 02

   Generate or paste the script

   Generate a draft, paste your own script, or edit the AI draft directly in the script editor.
3. 03

   Meet the minimum length

   The live builder requires enough script text before voice generation is available.
4. 04

   Preview the voice

   Create the voice preview from the exact script and settings you plan to render.

• Use Analyze presenter voice when you want Speakeasy to suggest a voice profile from the selected presenter.
• Edit the base voice description with tone, pace, accent hints, and delivery style.
• Use feedback when regenerating, such as slower, warmer, less salesy, or more direct.
• Tune speed, stability, similarity, and style before regenerating the preview.
• Use speaker boost when the voice should keep extra clarity and presence.
• Use subtle ambience only when background sound helps the video feel natural.

Upload sample

Choose a clean audio file. A single speaker, low background noise, and steady volume produce better clone previews.

Record sample

Record directly in the browser. The live UI shows recording time and asks for at least 15 seconds for clone samples.

Reference transcript

Optional but recommended. Paste what is said in the reference audio so the clone can align to the sample.

Style instruction

Optional guidance such as warm, conversational, steady pace, or more energetic.

Use final audio when the narration is already recorded, approved, and should not be regenerated. This mode accepts common audio formats, previews the selected file, uploads it, and skips the Script & Voice step.

• Supported upload types include MP3, WAV, M4A, AAC, OGG, FLAC, and WEBM.
• You can upload an audio file or record audio in the browser.
• After final audio is uploaded, the render uses that audio directly.
• Use this path for approved voiceover, podcast-style narration, or externally produced audio.

Out-of-date preview

If the script, voice feedback, or tuning settings change after a preview is generated, regenerate before rendering.

Save voice and continue

For designed voices, save the generated voice preview before moving to render planning.

Regenerate preview

Regenerate whenever the delivery does not match the final script or intended tone.

Direct audio

Direct final audio marks the voice step complete without generating a new voice preview.