Qtum.ai

Create visually stunning short dramas from a single prompt automatically with DreamZo, powered by the Qtum Router. No signups or credit card, privacy first.

[Qtum.ai @QtumRouter]

Qwen 3.8 27b price lowered on https://t.co/BhIbZx3SUM

2 cents for 1 million input tokens
24 cents for 1 million output

262k context, 16bit

[Qtum.ai @QtumRouter]

@hive_ai Appreciate the help @hive_ai

If you're curious what this image is and where it came from, it was a screenshot from the tutorial in DreamZo.

If it was an AI generated image, that would have been fine as well, because we're discussing a tool that generates AI videos!

[Qtum.ai @QtumRouter]

🧵1/12

DreamZo is a locally installed application that handles long-form AI video generation.

It's designed to help you plan a drama story, generate the visual characters, and keep the videos consistent between clips.

Use your own hardware, or plug it into the Qtum AI router.

[Qtum.ai @QtumRouter]

🧵12/12

https://t.co/nMh0uTiIwp

DreamZo is free, you can just download it and start building. If you have your own hardware you can generate the videos and speech. If you don't have any compute, you can simply plug it into the Qtum API router and pay for what you generate.

If you do have hardware, the video model DreamZo supports is MiniMax H3. For voice you need to use VoxCPM. Everything else is API hosted like the image generator and the LLM.

[Qtum.ai @QtumRouter]

🧵11/ 12

You can tell DreamZo to just "generate everything" if you want to try and one-shot a video. You can then go back and just regenerate the clips that you don't like.

Once you've got a video you're happy with, click export, and it will stitch it all together for you.

[Qtum.ai @QtumRouter]

@MiaAI_lab I bet there's people with a single DGX spark that want to add a second device simply and can't due to supply issues.

49 tokens per second is pretty unreal for such a capable model.

[Qtum.ai @QtumRouter]

@steipete @xeophon Accurate and powerful statement.

Anyone with a giant lobster claw hand is probably going to be good at patching things.

snip snip

[Qtum.ai @QtumRouter]

Laya Decision Model vs Human.

We used Fable 5.1 on https://t.co/bSyv5SnLxv to build a very simple snake game. You eat the apple and you get bigger. If you hit the wall or your own tail, you lose. Once the clock runs out, or a player dies, the score is tallied.

Laya won 11 times in a row until we gave up.

You can access Laya for 2.1 cents / 1 million input tokens on the Qtum AI Router. Output is free.

Lyrics were made with GPT 5.6, music by Seedance 2.5

[Qtum.ai @QtumRouter]

Same prompt but made with MiniMax-H3
Which one is better?

[Qtum.ai @QtumRouter]

Made with Wan3.0 at https://t.co/bSyv5SnLxv

Prompt: 15-second premium American rap music video, vertical 9:16. The explosive final chorus of a summer anthem. Photorealistic, polished cinematography, confident performances, strong visual rhythm.

LOCATION & CAST:
A modern hillside mansion with an infinity pool overlooking Los Angeles at sunset. An adult male rapper, early 30s, charismatic, athletic build, wearing an open ivory silk shirt, tailored swim shorts, a platinum chain, and dark sunglasses. Six stylish adult women, all in their mid-20s to mid-30s, wearing elegant bikinis and understated gold jewelry. A glamorous, relaxed pool party with natural interactions and effortless dancing.

MUSIC:
Original contemporary West Coast hip-hop at 100 BPM. Deep, clean sub-bass, punchy drums, a memorable dark synth motif. Confident male rap vocal, natural American accent, precise lip sync. Begin immediately on the chorus downbeat, no intro.

0.0–4.8 SECONDS — THE HERO SHOT:
Low-angle medium shot tracking backward as the rapper walks along the pool edge toward the camera, delivering:
“Whole crew up, let the skyline know /
Came from the ground, now the view all gold.”
Behind him, women dance casually to the beat beside the shimmering pool. Warm backlight, realistic skin texture, subtle lens flare. His gestures are controlled and rhythmic.

4.8–7.2 SECONDS — THE SPLASH:
Hard cut on the snare to a water-level shot. An adult woman in a black bikini rises out of the pool, sweeping her wet hair backward in one smooth motion. Backlit droplets arc through the frame in elegant slow motion. The rapper continues offscreen:
“Bass hits hard when the sun dips low.”

7.2–12.0 SECONDS — THE CHORUS PEAK:
Hard cut to a medium-wide shot of the rapper centered beside the pool, with the party moving naturally around him. The camera makes a smooth, shallow arc. He delivers directly to the lens:
“One more night, we ain't heading home.”
He lifts one hand on the final word as the bass lands. The vocal trails into a spacious echo, leaving room for the instrumental.

12.0–15.0 SECONDS — THE MONEY SHOT:
Cut to a wide crane shot, rising smoothly to reveal the entire illuminated infinity pool, mansion, palm trees, and glowing city beyond. The group continues moving to the beat as the instrumental peaks. End on the expansive cinematic view.

VISUAL DIRECTION:
Luxury editorial styling, warm amber highlights, rich cyan pool water, subtle film grain, realistic anatomy and water physics. Maintain the rapper's face, outfit, and jewelry across every cut. Fluid movement, readable compositions, edits precisely on the beat. No dialogue beyond the specified lyrics. No captions, text, logos, or watermarks.

[Qtum.ai @QtumRouter]

@TheAhmadOsman Can't wait for Qwen 3.8 27b to get updated.

[Qtum.ai @QtumRouter]

We have added an alternative to Jev called "Laya" on https://t.co/XXNDYDpJL4

$0.021 for 1 million input credits, output free.

We also added a new "Decision" panel.

https://t.co/1dhYWJjkEs

[Qtum.ai @QtumRouter]

@MiaAI_lab Would you mind sharing the memory used on the sparks?