Skip to main content
Text to Speech

Minimax Speech v2.8 Async

Generate high-quality natural speech audio from long text using Minimax Speech 2.8 async models with expressive voice options and English normalization for long-form input (up to 50K characters)

View details
Try it in Ampere

Inputs

Loading input fields...
Execution Steps

Loading workflow structure...

Loading curated examples...

Overview

Minimax Speech v2.8 Async turns long text into natural-sounding speech audio using MiniMax Speech 2.8 async models. Use it for long-form narration, training content, podcast segments, and script batches up to 50,000 characters when the standard v2.8 AI Tool's 10,000-character limit is too small.

Use cases

  • Generate narration for longer product walkthroughs, lessons, or podcast-style segments.
  • Turn a long campaign script into one audio file without splitting it into 10K-character chunks.
  • Test HD versus Turbo model choices for longer voiceover drafts.
  • Use pronunciation overrides for product names, acronyms, or brand terms in long scripts.

Input tips

  • Keep text under 50,000 characters and review it for pronunciation before generating audio.
  • Use a built-in MiniMax voice ID or an approved custom cloned voice ID.
  • Choose speech-2.8-hd for default quality or speech-2.8-turbo when speed matters.
  • Use <#x#> pause markers only when custom pauses are needed.
  • Set language_boost to a target language or auto for automatic detection.
  • Choose mp3, pcm, or flac output; set sample rate, bitrate, and mono/stereo channels for the handoff.
  • Use emotion, speed, volume, pitch, and sound effects sparingly for long scripts.

Expected output

The AI Tool returns one generated audio file with a downloadable URL, content type, file name, file size, async task ID, file ID, billed-character count, status metadata, and cost metadata. The shared Minimax speech output view renders an audio player and character-count metadata when available.

Caveats

  • Async runs can take longer because the task is created, polled, retrieved, and then saved.
  • Generated speech should be reviewed for pronunciation, pacing, emotion, and brand fit.
  • Invalid or unavailable system or custom voice IDs will fail validation.
  • The async output schema does not guarantee duration, sample rate, bitrate, channels, or word count metadata.
  • Pronunciation overrides and language boost help, but long scripts still need listening review.
  • Use the standard Minimax Speech v2.8 AI Tool for shorter scripts that do not need the 50K limit.

Related AI Tools