Audio model · MiniMax ·
MiniMax Speech 2.8 HD is an audio model by MiniMax. In Endless it costs from 1 credit per clip and is available on every plan.
MiniMax Speech 2.8 HD is MiniMax's high-fidelity text-to-speech model for voiceovers that need natural fillers, sound tags and emotion control.
MiniMax Speech 2.8 HD, released in January 2026, reads text aloud with the small sounds that make speech feel unscripted. You write tags such as (laughs), (sighs), (breath) or (chuckle) where they should happen, and mark pauses with a tag like <#1.5#> for a break of 1.5 seconds. The 2.8 release also models colloquial fillers like um and ah, so a read can hesitate the way people do.
Teams use it for ad voiceovers, explainer narration, character lines in games and podcast-style intros where a flat read would sound robotic. You can set the emotion (happy, calm, sad, angry, fearful, disgusted or surprised) and the speaking speed from 0.5x to 2x, then adjust the script until the delivery lands.
In Endless it sits next to the ElevenLabs voice models, so you can generate the same script with each in chat or on the canvas and keep the take you prefer, paid from the same workspace credits.
Start a chat in your workspace and ask for MiniMax Speech 2.8 HD by name.
Paste your script or describe the sound, its tone and its length.
Ask for another take or a different voice. Every clip is saved to your workspace assets.
by ElevenLabs
Expressive narration and character voices in 90+ languages
Price depends on your input
by ElevenLabs
Dramatic voice reads directed with audio tags, plus voice conversion
Price depends on your input
by ElevenLabs
Steady long-form narration in 29 languages with fine voice control
From 1 credit per clip
Start free, create a project, and bring your brand’s references together. Or book a demo to explore Endless with your team.