
Meta's FREE "Banana Killer" & My AI Video Tool (Also Free!)
Theoretically Media8 July 2026Watch on YouTube
What you'll learn
- You test how Meta's Muse Image model performs on prompt adherence, text rendering, editing and its visible thinking mode
- You see where Muse Image ranks against GPT Image 2 and Nano Banana 2, including the conclusion that the Banana survives
- You discover what the first Muse Video samples and the surprise Seedream 5.0 Pro release tell you about video models
- You learn how a free, self-built tool turns footage into an OpenPose skeleton and depth map for video control
- You see how that tool applies pose and depth control in Seedance, Runway and other AI video tools with multiple characters
Frequently asked questions
Where can you use Meta Muse Image for free?
Does the Banana still hold the top spot after the Muse Image release?
What does the free tool built in the video do?
Where can you download the free tool and the source code?
Topics
In this video
Read next
LUGN Security bewaakt grote terreinen met sensoren, drones en AI vanuit Heerlen
LUGN Security uit Heerlen combineert sensoren, autonome drones en een centraal AI-aangestuurd controlesysteem om grote terreinen en infrastructuur te beveiligen. Het bedrijf, opgericht in 2021, haalde €36,8 miljoen op in een Series B-ronde en werkt vanuit de Brightlands Smart Services Campus.
World Labs genereert duizenden gesimuleerde robottaken voor training
World Labs, opgericht door AI-onderzoeker Fei-Fei Li, presenteert een simulatie-engine die robotcontrollers trainen in virtuele omgevingen op basis van reële taken.
Meta rolt Muse Image AI-model uit via Instagram en WhatsApp
Meta's nieuwe Muse Image-model voor AI-beeldenmaking wordt geactiveerd in Meta AI, Instagram en WhatsApp, en volgt later in Facebook en Messenger.
Meta geeft korting voor delen van AI-gebruiksdata Muse Spark
Meta biedt een expliciete korting van gemiddeld circa 95 procent aan gebruikers van hun nieuwe Muse Spark-model die delen hoe ze het model gebruiken.
Hands-on: Meta's AI-agent Muse en het uitgeven van geld
Een test van Meta's AI-agent Muse beschrijft hoe de assistent taken oppakt, waaronder het doen van aankopen.
Meta lanceert AI-assistent Muse om achterstand in te halen
Meta introduceert Muse, een persoonlijke AI-assistent waarmee het bedrijf zijn positie in de AI-wedloop wil verbeteren via een miljardenstrategie.
Description from the channel
Meta just shipped Muse Image — a FREE "thinking" AI image model live now at meta.ai — and announced Muse Video, the first Meta video model that will actually release. Spoiler: the Banana survives (GPT Image 2 and Nano Banana 2 still hold the top spots), but Meta's model is free, real, and closer than expected. Plus: I built you a free AI video tool — precise pose + depth motion control for video-to-video in Seedance, Runway, and anywhere that takes a video reference. Download below. Source code included. ▶ Try Google Omni + the new AI Apps over on Artlist, who were kind enough to sponsor today's video: https://artlist.io/?artlist_aid=TheoreticallyMedia_4024&utm_source=a%5B%E2%80%A6%5Dum%3DTheoreticallyMedia_4024&utm_campaign=TheoreticallyMedia_4024 ▶ THE FREE POSE + DEPTH TOOL (put $0 in the box — it's yours): [https://theoreticallymedia.gumroad.com/l/zobqss ▶ Meta Muse: https://meta.ai/ In this one, I run Meta Muse Image through the full test gauntlet — prompt adherence, text rendering, image editing, and the visible thinking mode (including watching it talk itself out of a correct answer) — plus the Elder Scrolls 6 imagination test, then stack it against GPT Image 2 and Nano Banana 2 to see where it really ranks. We also break down the first Muse Video samples, ByteDance's surprise Seedream 5.0 Pro drop, and how the free tool turns any footage into an OpenPose skeleton + depth map (Video Depth Anything) that gives you real motion control over your AI video generations — multi-character tracking, in/out points, and the hardest video-to-video test I've ever run on the channel. CHAPTERS 0:00 Meta Actually Shipped Something 0:36 The Road Here: Make-A-Video, MovieGen & Vibes 1:40 Testing Muse Image (Free at meta.ai) 3:12 Thinking Mode: It Argues With Itself 4:02 The Gauntlet: Wine, Clocks & a Pelican 5:53 The Elder Scrolls 6 Test 6:53 The Text Test: Sharknado AI 7:22 Where Muse Gets Lazy 8:48 Verdict: Where Muse Image Really Ranks 9:13 Seedream 5.0 Pro Just Dropped (Quick Note) 9:37 Muse Video: What the Samples Tell Us 11:38 Sponsor: Artlist (Google Omni + AI Apps) 13:13 I Built You a Free Tool 14:09 How It Works: Pose + Depth Baking 15:58 Running It With Seedance (on Runway) 16:59 Multiple Characters, Cuts & In/Out Points 18:26 The Hardest Shot in AI Video 19:47 Get It Free (Gumroad + Source Code) MORE THEORETICALLY MEDIA Newsletter: https://theoreticallymedia.com/newsletter/ X/Twitter: https://x.com/TheoMediaAI Business inquiries: tim@smoothmedia.co #MetaAI #MuseImage #AIVideo