From Google DeepMind - A "genuine phase transition in the pace of scientific discovery"
Across our projects @GoogleDeepMind, we are witnessing a genuine phase transition in the pace of scientific discovery. You will see a glimpse of this in our releases in the next few weeks.
“An expert is someone who has made every mistake it is possible to make, in a narrow field”—Niels Bohr
Kling 4.0 is actually really good at restyling videos; better than Seedance 2.5 @Kling_ai SOUND ON
Amazing visuals!
LADIES AND GENTLEMEN: THE SURREAL.
SUPERNATURAL | A @surRIHL CUT
DIRECTED BY: @papadupapadu + @LoLoLumi88
MIDJOURNEY + KLING 4.0 VIA ADOBE FIREFLY + ELEVEN MUSIC V2.5 + ELEVENLABS SFX + MANUAL EDITING IN PREMIERE PRO + CLAUDE OPUS 5.5
Announcing LUMINA, the world’s first dedicated theatrical distribution program for AI-generated and AI-assisted films.
Another attempt at an extended conversation clip.
It is really hard for AI to do extended dialogue scene. I don't think there is some magic workflow.
I tried several different methods. If you know of a good, tried-and-true workflow, please drop it in the comments!
What we did
- Rebuilt the scene in a simpler basement set with an orange-moon window, an oil lamp and a frosted empty tank.
- Generated one master ElevenLabs dialogue for all 24 lines.
- Rescripted it with the reference clip's techniques and cut it to four generations of 26 to 28 s.
- Each used the black video carrying the dialogue, a timed transcript, the character sheets, the room sheet and a timestamped shot list.
- Assembled a 1:42 cut with a steady room tone, three very quiet foley hits and no music.
What worked
- One master track kept the voices consistent: pitch within about 3 Hz of the master on nearly every line.
- Putting timestamps for both lines and shots in the prompt pulled timing to within about half a second.
- The black video plus transcript worked well on G1 and G3.
- Prompt fixes for a duplicate Zoe, plain skin and an empty tank all landed.
What didn't
- The first seven-generation cut had a duplicate Zoe, cuts every 2 s, freezes after cuts, and foley that was too loud.
- Using the silent acted clip as a video reference gave weak lips, a copied, stiff performance and a drifting look.
- I misread a spectral measure as a voice-timbre shift. It was just s and z sounds.
- The master stage was boosting the quiet opening by 35 dB, and the room-tone loop swelled.
- G4's last line, "You.", is still not audible. The fix is a short extra generation for about $1.3, or a G4 re-roll for $3.60.
- Seedance sets its own speech pace by a second or two, so long holds need trimming in the edit.
David Michael Comfort retweeted
.@DreamLabLA is in the studio cooking up heat 📷
Real actors. Live sets that change on the spot. A crew working with AI in the room.
@FortuneMagazine went inside Dream Lab LA to see hybrid production in action. 📷
David Michael Comfort retweeted
NO ONE is ready for what is coming to video because of AI.
NO ONE.
It’s going to be insane.
Hollywood Has Flipped on GenAI, Says 'Avatar' VFX House Weta FX Chief: 'If You Don’t Have an AI Strategy, Don’t Bother'
Hollywood Has Flipped on GenAI, Says 'Avatar' VFX House Weta FX Chief: 'If You Don’t Have an AI Strategy, Don’t Bother' variety.com/2026/film/market…
What a great video!
our opening film for @mondaydotcom ELEVATE 2026 is here!
we teamed up with @invideoOfficial to see how far we could push Agent 2. turns out, pretty far.
huge thanks to Dan Brooks & Leif Huron for somehow letting us get away with this.
music by SonBon studio.
Here is a quick comparison between gpt-image-2.5 vs gpt-image-2.6.
TLDV - I used a variety of styles using references. In most cases it is a tie, but I did like some of the 2.6 images better.
What do you guys think?
How Grimes' Cameo in Theater-Bound AI Film ‘Gods Don’t Give Gifts’ Came Together (EXCLUSIVE) variety.com/2026/film/news/g…












