Skip to content
Nuvexa AI
Menu
Voice · PRACTICAL TUTORIAL

Prepare one, two or three characters for Nuvexa voice generation

A practical script workflow based on the current character-voice interface, with a measured text-count example.

Published by Nuvexa AI · Published · 3 min read

Start here

A practical script workflow based on the current character-voice interface, with a measured text-count example.

  • Your script with each dialogue line assigned to a speaker
  • Chosen voice for each character
  • A short pronunciation sample before the full script

Map the speakers before generating

For a dialogue, name the roles and assign every line to the intended voice. A narrator can introduce a scene while the other characters ask and answer. Two or three voices without a role map may sound like unexplained switches. Keep the character names consistent across revisions.

Test the important words

Generate a short excerpt containing names, brand wording, amounts and units. Listen before spending time on a long script. A voice setting alone does not establish correct pronunciation. If a name is important, have an appropriate reviewer confirm it. The current feature uses available catalogue voices rather than promising custom voice cloning.

Estimate duration from a stated pace

The public Word Counter counts “Check the file before you publish it.” as 7 words and 37 characters including spaces and punctuation. At the entered rate of 150 words per minute, it rounds the speech estimate up to 3 seconds. This output was checked using the current textStats implementation. It measures a text estimate, not generated audio duration.

Review continuity and downloads

Keep the character voice and rate choices with the script. When revising one line, compare the replacement with nearby lines for identity, pace and loudness. Play the downloaded MP3 outside the workspace and check its beginning and ending. A download control does not prove that every intended line is present.

Separate voice generation from transcription

Voice generation turns text into audio. Audio Transcription accepts a recording and returns editable text. Clipping edits supplied video. None of those names imply that video generation is active. Choose the feature matching the material you have, and check current availability before planning a larger production.

Review one short exchange before the full script

Begin with a few lines that include every intended speaker. Listen for correct speaker assignment, a natural pause and readable pronunciation. If one speaker sounds wrong, change that character’s voice or line rather than regenerating the whole script repeatedly. Keep the script and voice choices together in your notes. A word-based timing estimate helps plan length, but it cannot confirm the duration or pronunciation of the generated audio; listen to the actual file before downloading the final version.

Worked example and reusable input

Use the stated inputs to reproduce checked local outputs. Replace sample facts when applying a brief to your own work.

Checked text: Check the file before you publish it.
Words: 7
Characters: 37
Entered pace: 150 words/minute
Estimated speech: 3 seconds
Actual audio duration: not measured in this text-only check.

Key takeaways

  • Assign each line before generating.
  • Listen to a short pronunciation test.
  • Treat timing as an estimate until you measure the audio.