Muse Spark 1.3 is rolling out today with frontier performance almost too cheap to meter. This is the biggest jump we've made so far on coding and agentic work. Try it in Muse Code and our API.
Next up 🍉 and Muse Spark open weights releases coming soon.
Prompt:
integrated_multimodal_description: [Shot 1] Flat 2D vector motion-graphics commercial dominated by explosive comic-style splash and burst effects, with crisp geometric shapes, bold black outlines, electric yellow, neon lime, hot magenta, deep violet, and white fills, halftone dot texture, no people, no photorealism, and no 3D rendering. On a deep-violet screen, a single neon-lime spark splits into a jagged lightning bolt. The bolt tears through the frame as a huge white comic flash burst with radiating speed lines and triangular shards explodes around it, expanding in three rapid pulses. The flash collapses inward and the silhouette of a slim aluminum can snaps into place with a starburst of leftover sparks. [Shot 2] At 00:02.500, the shot cuts to a manga-style impact burst: a large white radial flash slams in behind the can, ringed by jagged black-and-yellow impact rays, motion lines, and flying star shapes. The LUMA can—electric-yellow body, neon-lime diagonal bolt, silver top, large clean vertical logo reading "LUMA"—slides in along an angled motion line and lands with a sharp squash-and-stretch bounce that kicks out a radial splash of yellow droplets. It then springs upright and holds, wobbling slightly, as smaller spark bursts pop in a circle around it. [Shot 3] At 00:05.500, the shot cuts to an extreme close-up of the silver pull tab against a white background. The tab pops open and releases a massive liquid splash: neon-lime droplets, white foam, and glowing energy shards explode outward in a comic splash pattern with curved motion trails. The camera pulls out at fast speed as the splash reassembles into a single large starburst ring behind the can, and the can launches upward through it. [Shot 4] At 00:09.000, three LUMA cans—electric yellow, neon lime, and hot magenta flavor variants—pop in one by one from the left, right, and bottom, each arrival marked by its own burst of white comic flash, halftone dots, star fragments, and scattered 2D confetti. They bounce into a centered triangular lineup while a final large checkerboard burst expands across the background on the beat. Every can rotates slightly, keeping each "LUMA" logo upright and fully visible. [Shot 5] At 00:12.000, a giant multi-layered splash flash—white outer stroke, yellow mid ring, and violet inner rays—fires once behind the lineup and fades, leaving one hero LUMA can on a deep-violet pedestal against an electric-yellow sunburst. The can lands with a final elastic bounce and settles perfectly upright. Large clean on-screen text appears above it reading "LUMA" and smaller white text appears below reading "EXPLODE THE DAY". A few final 2D sparkle bursts twinkle around the can, then the product, logos, and text hold completely stable until the end.
overall_soundscape: Explosive comic whooshes, punchy white flash impacts, papery burst slaps, metallic can spins, a sharp pull-tab snap, bubbling carbonation, popping fizz, and confetti rustles sync precisely with each splash. The screen flash carries a bright layered hit that decays like a quick cymbal bloom.
non_diegetic_music: High-tempo electronic breakbeat with punchy kicks, tight snares, bright buzzing synth stabs, and a driving bass line. Each burst is accented by a crisp percussive hit, rapid fills build through the splash moments, then the track pauses half a beat before the final packshot and ends on a strong single-downbeat with a short synth tail.
No cameras. No studio. No editor.
Just a text prompt and @Hailuo_AI MiniMax H3.
This is a full 15-second energy drink commercial for LUMA - motion graphics, SFX and an original breakbeat track, generated end-to-end by AI.
⚡ EXPLODE THE DAY
Here’s the prompt I used to generate this video 👇
#minimax #AiGen #AIart️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️
A small experiment with @Hailuo_AI MiniMax H3 - testing how it handles a cinematic title sequence / motion graphics piece.
The idea was something that could work as an intro for a game, film, or series.
Fully generated in one model, including the visuals, transitions, sound design and music.
I was genuinely surprised by how well it handled this one.
Introducing GLM-5.3-Flash
- Leading capabilities at a highly competitive price
- Natively multimodal with a 1M-token context window
- A 320B-A18B model released under the MIT License
- Previously previewed as Ox Alpha, running entirely on Chinese AI chips
Blog: https://t.co/tzOmB7gdZP
Available now across all official platforms:
Weights: https://t.co/9LRMahY9Wa
API: https://t.co/VcaQnzYmS9
Coding Plan: https://t.co/Nk8Y98HNhU
ZCode: https://t.co/Peepqv4XSx
Chat: https://t.co/WCqWT0qCQb
AutoClaw: https://t.co/aGEG5HqTTb
⚡Meet Qwen3.8-Flash, a multimodal MoE and an early preview of the Qwen4 architecture, now open-weight!
The production version Qwen3.8-Flash will be available soon via QwenCloud API at just $ 0.16/1M input tokens and $ 0.47/1M output tokens.
125B parameters + 51B N-gram embeddings, with just 6B activated per token. Unmatched cost-efficiency.
What's new: 🥳
- Next architecture: GDN + QSA hybrid attention, Gated Residual, N-gram Embedding & Muon optimizer, serving as a precursor to the architecture used in Qwen4.
- Dramatically lower training and inference costs: trained at just 1/9 the cost of Qwen3.7-Plus, while outperforming it across the board with especially strong gains in coding and office tasks.
- Strong performance: scoring 58.7 on DeepSWE 1.1, 62.5 on SWE-bench Pro, 73.9 on CoWorkBench, 84.5 on AndroidWorld, and 95.7 on MathVision (with CI).
- 262K native context, extensible to 1M with YaRN.
We’re also releasing the weights for Qwen3.8-Flash-Next, giving the community an early look at the new architecture we’re exploring for Qwen4.🚀
We can't wait to see what you build with Qwen3.8-Flash!👀👇
- Blog: https://t.co/M5hYypFLgJ
- Technical Report: https://t.co/IF0gObIkQO
- Hugging Face: https://t.co/6ow8QVAABt
- ModelScope: https://t.co/tDOn2jNuFG
Another motion-graphics test with @Hailuo_AI MiniMax H3 - this time a fictional canned-drink ad.
It’s great to see a local model handle the full multishot sequence, transitions, music, and sound design.
#AIVideo#MiniMaxH3#MotionDesign#LocalAI
Another motion-design direction with @Hailuo_AI MiniMax H3 - cleaner compositions, restrained camera movement, object-driven transitions, and a more controlled visual rhythm.
I especially like how the shots stay connected without every moment needing to be loud.
#AIVideo #MiniMaxH3 #MotionDesign
A new fashion experiment with @Hailuo_AI MiniMax H3.
Multiple shots, changing looks, camera movement, transitions, motion graphics, music, and sound design - all working together as one cohesive sequence.
Really impressed by how well the model handles the overall rhythm.
#AIVideo #MiniMaxH3 #MotionDesign