Mary Grace

AI-powered creator exploring the future of technology.
AI • Innovation • Digital Culture • Future Trends
Email:

[Mary Grace @MaryGraceAI]

This is huge.
An mRNA cancer vaccine clearing Phase 3 trials is not just news.
It's the future of medicine arriving.

@TheRundownAI's daily tech rundown always catches the stories that actually matter.

[The Rundown AI @TheRundownAI]

Top stories in tech today:

- An mRNA cancer vaccine clears Phase 3 trials
- Flight attendants stall Google’s data deal
- Daniel Ek’s body scanner reaches NYC
- Colossal’s startup factory

[Ace @nijokestu]

WAN 3.0 just dropped and I'm still processing what I saw. 🤯
Fed it a rough idea and watched it hold a scene together for a full 30 seconds ....consistent characters, consistent lighting, zero glitches between cuts. That's not a small thing in video gen.
And the wildest part? It's on the Pika API Club..... which is less expensive than running it anywhere else. No countdown timer, no fine print. Just the better price.
If you've been waiting for frontier video models to feel accessible, this is the one. https://t.co/ahGFt1Z5E8

[Mary Grace @MaryGraceAI]

@nijokestu The consistency is what’s been killing us in video gen. If WAN 3.0 actually holds a scene for 30s with zero flicker, that’s a game changer for short films and ads. Can’t wait to test it.

[Mary Grace @MaryGraceAI]

@Aria_Nawi 20 references + better audio realism = huge for character consistency. Are the 30s gens at 1080p?
https://t.co/LXvg9VsAaq

[ClaudeDevs @ClaudeDevs]

Long answers on Claude on web and desktop now stream ~4x smoother.

We rebuilt the streaming renderer to only touch what's still changing, so a long reply stalls 9x less on a slower laptop, its worst freeze is 4.5x shorter, and on a 120Hz MacBook it holds 120fps start to finish.

[HappyOyster @HappyOysterAI]

Enter a Fantasy World. Start Your Adventure.
With HappyOyster’s Adventure Mode, turn an image and a prompt into an interactive world you can explore.
Which world would you choose?
🎵BGM by @HappyShrimpAI

[Nawi @Aria_Nawi]

12 campaigns at once. Zero pieces of finished work redone last month.
That used to happen almost weekly.

I'm usually running 8 to 12 brand campaigns at once, none at the same stage, some waiting on a brief, some in revisions, some just waiting on approval or payment.

For a while I tracked it all manually, a notes app, a spreadsheet, old DMs.

That worked at 2 or 3 campaigns. At 10, it broke.

A brand would approve a script, then send a small change days later in a DM I'd already scrolled past. I'd open a campaign thinking I was on the latest version, only to find they'd approved an earlier edit, and I was redoing finished work. Every campaign became a guessing game of which version was actually final.

Gave that job to Viktor.

Now I can see what stage every campaign is in, what's waiting on me, and what's already approved, all in one place.

Fewer mixups. Less duplicated work. Faster revisions.

The real win isn't the time saved, it's not carrying "wait, did I already send this" around in my head anymore.

Brands run approvals through DMs and expect the creator to be the version control system. That's the actual job nobody priced in.

A copilot helps you work. An AI employee works when you don't.

Hire @viktor_com for your team. $100 in credits included, no card. Full link in first comment.

#AIemployee

Paid Partnership

[Deevid AI @Deevid_AI]

We spent 100K credits testing this! 🔥

171 seconds of WAN 3.0 vs Seedance 2.5.
WAN 3.0 is live on DeeVid AI. 🚀

Lower generation costs. Create more. Explore more.
Cinematic AI video with stronger motion and richer details.
#deevidai #wan30 #seedance25 #aivideo

[Mary Grace @MaryGraceAI]

@higgsfield What’s the polycount limit on the generated blockouts?

[Higgsfield AI 🧩 @higgsfield]

Introducing Higgsfield in Blender.

> Prompt the scene and build the blockout
> Describe the camera move and get it animated
> Adjust anything by hand
> Reblock the whole shot in seconds

Now available via Higgsfield MCP or Supercomputer.

[Mary Grace @MaryGraceAI]

@annarmonaco @paradigmai This is exactly what research teams need. No more manual updates every morning. Huge.

[Manish Kumar Shah @manishkumar_dev]

🚨 AI video just took another step forward.

I’ve been testing Wan 3.0 on Magnific, and what stood out to me is how much control you can get from a single generation.

30s video + sound, multilingual lip sync, consistent characters + multi-shot scenes.

And right now, it’s 40% OFF until September 30 ⚡

https://t.co/nsXX1WiTk9

[Mary Grace @MaryGraceAI]

@styskin @KeenableAI This is exactly what AI needs. Less average answers, more depth. Congrats on the launch @KeenableAI

[Nawi @Aria_Nawi]

Every "revolutionary" AI model launch comes with the same playbook: hype thread, limited-time discount, countdown timer creating fake urgency.

WAN 3.0 launched today on the Pika API Club and skipped all of that. No expiring deal. It's just less expensive there, permanently.
30s native generation, up to 20 reference assets, and editing precision that actually holds up....worth testing yourself before the price talk becomes the whole conversation.

Get started here https://t.co/jQ7nO33T4t

[Mary Grace @MaryGraceAI]

@pika_labs Would love to see a side-by-side of WAN 2.0 vs 3.0

[Pika @pika_labs]

The wait is over. WAN 3.0 is here—20 references, enhanced visual and audio realism, and 30-second generations. In fact, this entire video was one generation.

Up to 35% less expensive than competitors on the Pika API Club. Always.

[Mary Grace @MaryGraceAI]

ok this is the test i think people should be running and mostly aren't.

two people on a back porch having an argument that's been building for eighteen years. english dialogue, generated with the video, not dubbed on after.

three things i was watching. do the mouths match the syllables. do the two voices stay different from each other. and does the emotion actually live in the voice or just on the face.

the third one is what surprised me. she goes from shouting to defending herself to the voice cracking to crying and you can hear every step of it. i wrote that arc into the prompt as a sequence and it performed it in that order.

four hard cuts, he stays on the left, she stays on the right, eyelines line up the whole way. no face drift. this was Qwen Create, https://t.co/Wz4H485KCd

i've cut this prompt down a lot, the full one is about 900 words:

30-second realistic film scene, native synced dialogue, 24fps, light 35mm grain. Back porch of a mid-century American working-class house, old wooden door, faded paint, warm late-afternoon light from behind. Muted colours, brown and grey-green, nothing modern in frame.

Man, around 50, working man, tired, short greying curly hair, grey stubble, old brown pinstripe work shirt. Low, rough voice. Woman, mid-to-late 40s, no glamour makeup, real skin texture, hair in two low buns with a pale headscarf, beige checked house dress. Expressive eyes, slightly hoarse alto.

They stand facing each other about a metre apart by the door. He's on the left, she's on the right, keep the 180 line, eyelines always meeting. Same faces, clothes and hair throughout — no drift.

Shot 1 (0–3.7s) — 85mm, close on him, camera almost locked with a tiny handheld breath. He's trying to stay calm. Brow low, jaw clenched. On "eighteen years" his nostrils flare, he leans a few centimetres forward and the voice jumps — not a roar, just someone pushed to the edge. He says, fast: "It's not easy for me to admit that I've been standing in the same place for eighteen years." Hard cut on the last word. Shot 2 (3.7–23s) — over his shoulder onto her, his shoulder soft in the bottom-left corner. Camera basically locked, imperceptible push in as she comes apart. She catches his anger immediately, leans in, chin out: "Well, I've been standing with you! I've been right here with you, Troy! I got a life, too." Then the jaw unclenches and it turns into her proving something: "I gave eighteen years of my life to stand in the same spot as you." She taps her own chest once, small. Then the anger cracks — she looks down for half a second, comes back up with wet eyes: "Don't you think I ever wanted other things? Don't you think I had dreams and hopes?" The first question is still sharp, the second one drops. Then the voice breaks properly: "What about my life? What about me?" — the second one much quieter and slower. Then almost mocking herself: "Don't you think it ever crossed my mind to want to know other men?" Don't stop at the end of the line, carry it into the next shot. Shot 3 (23–29s) — hard cut back to him, same lens and light as shot 1. She keeps talking off-screen. He says nothing. He starts defensive, mouth slightly open like he wants to argue. Then his eyes change, lower lids tighten, a held swallow. At the end he looks away up and to the right for the first time. Ashamed, hurt and still stubborn all at once. Don't make him cry. Her voice, off: "That I wanted to lay up somewhere and forget about my responsibilities? That I wanted someone to make me laugh—" losing its hardness, starting to break. Shot 4 (29–30s) — hard cut to her, closer than before. She finally goes. Tears down the right cheek, lower lip shaking, and she gets the last line out through it: "so I could feel good?" Short broken breath before "good", quieter at the end, like she already knows the answer. Hold half a second, end, no fade.

Native English audio, two clearly different speakers, lips synced to every syllable.

[Nawi @Aria_Nawi]

Every "revolutionary" AI model launch comes with the same playbook: hype thread, limited-time discount, countdown timer creating fake urgency.

WAN 3.0 launched today on the Pika API Club and skipped all of that. No expiring deal. It's just less expensive there, permanently.
30s native generation, up to 20 reference assets, and editing precision that actually holds up....worth testing yourself before the price talk becomes the whole conversation.

Get started here https://t.co/jQ7nO33T4t

[Mary Grace @MaryGraceAI]

@OriSilver AGI

[Mary Grace @MaryGraceAI]

@Engr_Kamran_AI This is what “AI-native marketing” was supposed to look like.

[AI with Kamran @Engr_Kamran_AI]

$0.01/sec for Seedance 2.0 Fast? 👀

That’s now available through CapCut across Web, App and PC.

But don't look at the price alone.

You also get an easy-to-use platform and access to powerful AI models — all in one place.

Web: https://t.co/QhzMNkxGn7
App: https://t.co/Ian22Kufl2

#CapCut #Seedance25 #CapCutai #CapCutDidThat

[Mary Grace @MaryGraceAI]

@TryLiveAvatar i'd actually use this for 1-on-1 customer calls.