me20
All news
/
News.Issue 11 ·

AI news: Runway adds sound, Voxtral 3 voices, Helix World

AI creative tools news, 9–22 October: sound effects and voiceovers in Runway, Mistral's Voxtral 3 voices and Helix World, an image you can walk into.

By Olena, me20 Studio · 2 min read

Stylised 3D portrait of a pale character with tousled platinum hair, golden eyes and black branch-like tattoos on the hand and neck, wearing rings and a floral jacket
stylised 3D character portrait, pale androgynous figure with tousled platinum hair and golden eyes, black branch tattoos on hand and neck, silver rings and chain, cream shirt and black floral jacket, soft studio light on a warm taupe background, moody cinematic close-up --ar 16:9

The biggest AI creative news of the last two weeks is about sound. Runway can now add sound effects and voiceovers through its agent, Mistral launched Voxtral 3 with expressive voices in 35+ languages, and a research team showed Helix World, which turns one image into a scene you can walk through with matching audio.

Runway's agent now adds sound effects and voiceovers

Runway added two audio models to Runway MCP for paid plans. Fish Audio Drama 3 (14 October) turns a script into an expressive voiceover with seven voices and cues for whispers, laughter and dramatic moments. Mirelo SFX 2.0 (15 October) takes a clip and a description, from footsteps to ambience, and returns a separate sound-effect track for your edit.

Background removal and fast reference images

Also in Runway MCP for paid plans: BiRefNet v2 (15 October) removes backgrounds automatically by detecting the foreground on its own — Runway names product shots, illustrations and assets for composites as the use. Seedream 5.0 Flash (12 October) generates 1K or 2K images from text or from up to 10 reference images.

Mistral Voxtral 3: expressive voices in 35+ languages

Mistral AI released Voxtral 3, a family of speech models and a voice agent SDK. The text-to-speech model, Voxtral Mini TTS 3, adjusts emotion, pace and emphasis to the context and supports 35+ languages and 50+ dialects, including Polish and Ukrainian. It costs $13 per million characters, and Pro Voice can adapt a speaker's style from about 30 minutes of audio. It runs on Mistral's API — with European, US and global endpoints — or on your own servers.

Helix World: one image becomes a world with sound

Noiz AI and the Hong Kong University of Science and Technology published Helix World on 12 October. From a starting image and an optional prompt, it generates a scene you move through with the keyboard, with stereo sound that follows the picture — 768×512 at 24 frames per second on a single data-centre GPU. It is a research preview with a public demo and code, not a finished product.

Want to put the new audio tools to work? See how to localise video ads with AI dubbing and our AI video work. Planning a campaign with sound? Tell us about it.

P.S. The caption under the cover image is a ready-to-use prompt. Copy it into Midjourney, Higgsfield or any image generator and make your own version.

Also in News