Олександр Цюцьмаць

AI | Claude | ChaGPT

[Олександр Цюцьмаць @iwannapods]

Booked a massage… and got the lion 🦁💆‍♀️

Most professional masseur alive 😳 Sound on 🔊

Would you dare? 👇

#lion #massage #viral

[Олександр Цюцьмаць @iwannapods]

Lionel Messi or Sashka?

[Олександр Цюцьмаць @iwannapods]

Workflow tip most people skip:

Never settle for your first generation.

Batch 6–8 variations of the same prompt, change ONE variable each time — light, lens, or motion.

Then cut, don't create.

The magic clip is usually #5, not #1. Generate wide, select hard.

[Олександр Цюцьмаць @iwannapods]

Reminder for AI video creators: Sora's API shuts down September 24.

OpenAI already killed the Sora app in April.

Right as Grok Imagine, Kling 3.0 and Veo 3.1 hit their best form.

The tool you build your workflow on matters. Don't get stranded on a dead one.

[Олександр Цюцьмаць @iwannapods]

Gemini Omni Flash can now edit your videos 🎬

No complex editor — just describe what you want (cut, trim, remove, transition) and the AI assembles the final clip for you.

Hours of editing, now done in minutes.

#geminiomniflash #ai #videoediting

[Олександр Цюцьмаць @iwannapods]

Best explanation of AI agents I've seen this week — from Jensen Huang:

The model thinks.
The harness gives it form.
Tools and skills act.
The runtime executes.

A worker in a workshop.

If you build with AI, internalize this. It's the whole agent stack in one image.

[Олександр Цюцьмаць @iwannapods]

4. Add sound. 5. Cut to the beat.

Drop in voice or music, trim every dead frame, end on the payoff.

That's it — trend to finished clip in 30 min.

Repeat daily and you'll out-ship 99% of creators. Follow for the next one.

[Олександр Цюцьмаць @iwannapods]

How to turn any trend into a finished AI video in 30 minutes.

No editing skills. No camera.

Here's the exact 5-step workflow I use 👇

[Олександр Цюцьмаць @iwannapods]

3. Animate with a multi-shot model.

Feed the frame into Kling 3.0 or Seedance 2.0. Describe camera movement, not the subject — 'slow push-in, 85mm, shallow focus.' Generate 4–6 takes, keep the best.

[Олександр Цюцьмаць @iwannapods]

Kling just shipped 3.0 Turbo.

The upgraded Omni mode does multi-shot sequences with one shared audio timeline — native dialogue in 5 languages, synced.

We're past 'AI clip.' This is AI scene direction.

Single-shot generators are about to feel ancient.

[Олександр Цюцьмаць @iwannapods]

Hot take: the bottleneck in AI creation was never the tools.

Models can already out-shoot most of us.

The edge is taste — knowing what's worth making and when to stop.

Prompts are cheap. Judgment is the moat.

Build taste, not just a tool stack.

[Олександр Цюцьмаць @iwannapods]

The market has a clear favorite in AI.

Polymarket now gives Anthropic 95% odds of holding the best model — vs Google at 3% and OpenAI at 2%.

Meanwhile Anthropic is paying $750k+ for LLM architects.

Whoever wins the model race wins the talent race too.

[Олександр Цюцьмаць @iwannapods]

There's a new #1 in AI video.

Grok Imagine Video 1.5 took the top spot on the Image-to-Video Arena — beating Sora 2, Veo 3.1, Seedance 2.0 and Kling in blind tests.

xAI went from no video model to first place fast.

Lesson: the leaderboard flips monthly. Stay flexible.

[Олександр Цюцьмаць @iwannapods]

A full presentation in 1 minute from a single topic 🤯

Gamma App does it all:
📊 Builds slides from your prompt
🎨 Picks design & structure
📤 Exports to PowerPoint, PDF, Google Slides

Free tier ≈ 10 presentations.

#ai #gamma #presentations

[Олександр Цюцьмаць @iwannapods]

Do you need 1000 followers ? Just reply = Hello" Lets follow you instantly 💛📷 💛

[Олександр Цюцьмаць @iwannapods]

Seedance 2.0 VFX
Promt:

@image_1 is the first frame and the environment — the young woman already standing in the white gothic room. @image_2 is the character reference — lock her face, hair and outfit to this image across the whole shot.

A bleached bone-white gothic hall — chalk-white walls, floor, throne, bookshelf, candelabra, everything monochrome white. The young woman is the only source of color in the frame: green/teal/blonde streaked hair, smoky charcoal eyes, burgundy lips, oversized dark hooded vest. She walks slowly through the empty white space. At set moments the image breaks: [VFX: hard digital glitch — frame stutter, RGB chromatic channel split, horizontal datamosh tearing, scan-line displacement — she instantly teleports to a different spot in the room, leaving a brief colored ghost-trail smeared across the white]. Between glitches she keeps walking calmly, then snaps to a new position on the next hit.

[00:00-00:02] She walks across the white hall from the left. Locked wide shot, 35mm.
[00:02-00:03] [VFX glitch hit] — she jumps beside the white throne, RGB split, colored trail.
[00:03-00:05] She walks slowly toward center, calm.
[00:05-00:06] [VFX glitch hit] — she snaps to the white bookshelf, datamosh tearing.
[00:06-00:08] She drifts back toward center, the only color in the white void.
[00:08-00:09] [VFX glitch burst] — she flickers rapidly between two positions, stacked colored echoes, then settles facing camera.

Camera: single locked wide shot, the camera itself jitters one frame on each glitch hit, otherwise still — so the teleporting reads as her, not the camera. Lighting: soft shadowless white-room light, flat and even.

Same character across every glitch: green/teal/blonde streaked hair, smoky eyes, burgundy lips, oversized dark hooded vest. Identity locked.

Style: ARRI Alexa base with intentional digital glitch art — crisp chromatic aberration, clean datamosh, scan-line displacement. 4K, sharp clarity, high-contrast white environment, she is the only color, glitch artifacts crisp and deliberate.

[Олександр Цюцьмаць @iwannapods]

An open model just caught the frontier.

GLM-5.2 — 744B params, MIT-licensed — matches Claude Opus 4.8 and beats GPT-5.5. At $1.40 per million tokens.

The open vs closed gap didn't shrink. It collapsed.

Frontier power at near-zero cost is here.

[Олександр Цюцьмаць @iwannapods]

The AI tools I actually use every day 🛠️

🎬 Kling 3.0 — 4K video with sound & lip-sync
🖼️ ChatGPT Image-2 — images for any idea
🎵 Suno — music from scratch

These 3 cover 90% of my content.

#ai #kling #suno