Director • CD • Writer | Sundance • NBCU • MTV • Super Bowl | UCB improv. Exploring AI storytelling tools. CPP Invideo, Kinovi, Higgsfield 🐻⬇️
Joined July 2009
- Tweets 6,605
- Following 2,325
- Followers 5,764
- Likes 7,078
76 Photos and videos
Pinned Tweet
Jun 10
Could this be in our future?
'The Tesla Pill' satirizes the convergence of AI and biotech.
Maybe someone should ask @elonmusk if this is possible ?
- or maybe not! 🤣
Shoutout to @kinovi_ai for help producing several shots for this redux.
As always I'm happy to share details on the process / prompts - just DM.
3
1
8
1,667
Jun 11
Another banger lol
1
72
Jun 10
Agent One @invideoOfficial is proving to be v useful.
Jun 9
This is how the biggest fashion brands use Agent One to build end-to-end campaigns for their new product lines and seasons.
We spent the last two weeks and over $5,000 figuring out the right way to do this - testing models, testing workflows, finding what actually holds across a full fashion campaign until we landed on a repeatable workflow that worked.
The full breakdown is in the video and the thread below.
1
1
69
Norrisimo retweeted
Jun 5
Respectfully, given how the training phase actually works and how the models are actually trained, the whole "theft" argument isn't accurate.
As far as VFX, AI in all its forms, including generative AI, has been used in VFX for years. It was being used long before the whole public outrage/media fear mongering thing started. Crowd sims, rotoscoping, object removal, clean plate creation, digital set extension, de-aging, face replacement and many other elements are AI assisted. It's one of the reasons that you don't see a ton of VFX artists freaking out about it. We know what it does well, we know what it does poorly, and we know where it can be best employed in the post-production pipeline.
2
1
13
313
Jun 5
Why burn credits on single clips when you can batch a whole scene? This is a test of pt 1. of our comedic micro series called 'Breaking History':
Jax and Kira receive assignments from REG-47 — a smug British AI voice — via a glowing closet portal in their Brooklyn apartment. Kira and Jax are 'revisionists' on contract to fix historical events. The stakes: Fail a mission and reality rewrites around you.
Next we'll see Jax and Kira in 1865 trying to prevent the assassination.
Video output produced via Zap Boards (currently in beta). This tool is intended to enable creators to drag and drop character sheets, show bibles, production design, storyboards, fields for camera control, dialogue, tagging audio master prompts for maximum control over outputs (up to :15s). Hooked into the best models via API.
Two of my fav features are the toggle for white grid lines over faces to get past the Seedance filters, and an export button that stacks the boards or any ref image / video on top of the video generation.
The tool is a chrome extension. Looking forward to sharing more as we continue testing.
Shoutout to @invideoOfficial agent for helping to produce the character sheets and image iterations. I ran out of credits to keep going but look forward to further testing of their micro-series capabilities.
p.s. Josie is my actual dog's name and yes she is a Bernadoodle.
167
Norrisimo retweeted
Jun 3
For a Microdrama studio, the game isn't only a great script. It's shipping enough of them, fast enough, to find your winner as soon as possible.
So we built a workflow inside Agent One that makes the production pipeline 5X faster - whether you're running a studio or an agency taking on series projects.
Time, cost, exact steps - all in the video and the thread.
42
40
218
32,358
Jun 2
We are not talking enough about duplicate frames being produced in AI vid generations that are supposedly 24fps- scrub through your clips, see those duplicate frames ? Now you know why your clip looks jittery.
1
47
May 18
Loving this approach for batch creating multiple clips in a 15s time frame using Seedance 2 Omni: Generate the storyboard via ChatGPT2 or NanoBanana Pro→ add master prompt, render the entire video → The result: Production ready boards a :15s @TrustTheBum branded 'Day in the Life of Sonny' video set in photorealistic environments, all rendered in one pipeline. Created w @kinovi_ai ! 🙏⚡️
1
167
Norrisimo retweeted
May 11
Discover Dreamina Octo 🐙
A new kind of creative relationship is taking shape — amplified by Dreamina Seedance 2.0.
It doesn’t just respond to prompts-
it understands your intent, expands your ideas,and creates alongside you.
Sharing the same screen, the same rhythm, the same vibe-
Octo moves with your thinking, turning sparks into something real.
Where thoughts are seen,and ideas are carried further, amplified at every step.
Coming soon!
-->Join the beta — signup form in the comments.
-->RT Comment within 12h to get 1000 extra credits via DM! First 200 users ONLY!
224
203
676
6,391,688
Apr 25
Banger workflow to create fast coverage using Seedance 2.0 Omni reference mode:
Rather than creating your coverage shots one by one in Nanobanana pro you can prompt for video in SD tell it to use first image as the starting point to create ten or so new camera angles of that same moment. Describe the characters, action, camera / lens type etc. Then list the angles 1-10 such as MS, CU, MCU, low angles, pans, tilts etc.
Keep in mind time you have for each shot and reduce or increase runtime as needed. (10 shots averaging out over 15s cut pretty quickly, so you may have to reduce runtime if you're capturing dialogue).
Make a character sheet and use as a reference for the shots you've specified. Grid them if you are getting face detection flags. This grid maker by @djakakajdis is awesome grid-adder-for-seedance-2-0.…
This way you can get 1:1 character consistency and a bunch of new angles that you can use right away. Or you can extract still frames, upres and use your favorite stills to generate new clips. **Also very useful if you just want the same dialogue line spoken in different angles.
Prompt below, it's not all perfect but for this use case there's some good options for extraction.
Made on @kinovi_ai with Seedance2.0Pro.
credits: Phil Langer @CuriousRefuge
1
1
4
341
Apr 25
Use @Image1 1 as the starting point and inspiration to create ten new camera angles shot on ARRI Alexa 35 with Arri Master Prime lenses, matching the look of @Image1 , subtle camera movement, interesting shots, always focused on the subject, his action is 1. looking confused at the lottery ticket 2. realizing he won the lottery. Subject: Match @Image2 for young handsome man's exact likeness. "rendering_priority": "ULTRA_PHOTOREALISTIC_CLEAN_FACE: Youthful clear skin, subtle freckles, render subtle skin pores, light stubble, organic grain, micro-creases, subtle natural moisture. Absolutely zero digital glitching, zero vertical lines, and zero compression artifacts on the face. Skin must be smooth and clear but textured with 100% anatomical accuracy especially on the face edges." After realizing he won, he repeats this dialogue in each shot, yelling out happily, naturally, "I'm rich! I won! Hell yes!," in several different ways. Here is list of shots: 1. Medium shot head on, 2. CU head on, 3. 3/4 CU left side, 4. 3/4 CU right side, 5. cu lottery ticket in his hand, ticket matching @Image3 camera moving across from 3/4 profile cu framing to frontal view cu framing, 6. his pov close-up on the lottery ticket matching @Image3 in same hand as in @Image1 7. dirty over the shoulder shot focused on the lottery ticket using @image5 with his face out of focus in foreground on right. 8. overhead shot hawk pov use @Image4 for reference, woman matching @Image5 present in this shot leaning against the door of her jeep facing him, her bottom half not visible as jeep is covering that part of her up. 9. low angle medium shot, 10. extreme overhead wide shot use @Image4 as reference, woman @Image5 small in frame leaning against the door of her jeep facing him, her bottom half not visible as jeep is covering her bottom half. "AUDIO REQUIREMENT": Zero music.
1
112
Apr 10
AI filmmakers are posting so much amazing work. Inspirational! Bravo
3
104
Apr 3
This
Apr 2
I've been testing Google Stitch 2.0 for a week now and it's genuinely replacing Figma in my workflow.
→ Infinite canvas where you drop references, screenshots, and everything in one place
→ Voice mode that feels like talking to a colleague about your design
→ Auto-generates the next screens by predicting what you haven't built yet
→ DESIGN(.)md gives you a portable design system in markdown
→ MCP server that pipes designs straight into Claude Code
If you're still designing MVPs in Figma from scratch, you're wasting hours.
Stitch into Claude Code via MCP is the pipeline that kills "AI slop" for good.
82
Apr 1
Currently experimenting with a new high fidelity method to produce amazing high resolution photorealistic images. Enjoying the speed and ease of the @invideoOfficial platform to execute these. Happy to share tips, just comment 'FLOW' or dm.
1
91
Apr 1
👀
Mar 31
most lipsync models break the moment a scene gets complicated
sync-3 was built for complicated
multiple speakers. low light. overlapping dialogue. one take.
1
75
