Single-Pass Multi-Shot: Directing Cuts Inside One Generation
The unit of AI video generation is no longer one shot. A look at models cutting between angles mid-generation, and why the third cut is usually where it breaks.
7 min read
Sketch to Finish: Turning Rough Marks Into Production Artwork
Sketch conditioning is underused because people draw too much. What a sketch-to-image model actually reads, and the minimum sketch that gets it there.
7 min read
Small and Fast: When a 6B Image Model Beats a 32B One
Parameter count predicts a ceiling, not a workflow. Why a 6B model can beat a 32B one in practice, and where the handoff from cheap to expensive belongs.
7 min read
The Sora API Sunset Is September 24: A Migration Plan
OpenAI shuts the Sora 2 API on September 24, 2026. A migration plan for Sora-shaped video workflows, mapped to Flux 3, Kling 3 Turbo, Grok Imagine, and VEO 3.1.
8 min read
Sound Design Before Music: Building a Bed From Effects
Most creators reach for a music track when a clip feels empty. Often what it actually needs is a room — build the ambience first and music gets easier.
7 min read
Spark Ads and Whitelisting: How Brands Turn Creator Posts Into Paid Media
How TikTok Spark Ads and Meta Partnership Ads let brands run paid media through a creator's own post — the authorization mechanics and the refresh problem.
7 min read
Spotify Canvas Loops: Eight Seconds That Sell a Track
Spotify Canvas is 3-8 seconds, vertical, and loops on repeat. A clip generated to loop beats one trimmed from footage that was never built to.
7 min read
Steam Capsule Art to Spec: Dimensions, Legibility, Rejections
Steam's capsule art has fixed dimensions and a legibility bar most AI key art fails at 120x45 — the size that actually shows up in a player's library.
8 min read
Stem Separation and What You Do With Each Half
Splitting a track into vocal and instrumental stems is the easy part. What each half is actually for — and a different tool entirely — is not.
7 min read
Style References vs Fine-Tunes: Which One Locks Your Look
Style references are instant and cheap but drift over a campaign; a fine-tune is durable but needs curated data. Where the crossover actually sits.
7 min read
Subtitle or Dub: Choosing per Market
Dubbing every market feels like the premium choice. The coverage math and the accessibility law both say subtitle broadly and dub selectively instead.
7 min read
Subtitles, Text Overlays and Timed Overlays Are Not the Same Tool
Ask for 'captions' and you might get a transcript of your voiceover, a supplied hook line, or a burned-in overlay that never touches the audio.
8 min read
Tagged, Not Removed: What Streaming AI Music Labels Change
Tagging, de-recommending, and removing AI music are three different platform responses. Deezer's own numbers show which one the industry is converging on.
7 min read
Text Baked Into the Image vs Text Stamped Over It
Two different ways to put words on a slideshow image — rendered inside the picture or layered on top — and the double-render mistake that breaks both.
7 min read
Text-Thread Scenes: Turning a Conversation Into Watchable Video
How to turn a text message exchange into a paced video scene: reveal timing, a reaction cutaway, and two different ways to voice the conversation.
7 min read
Text to Vector: When You Need an SVG, Not a Raster
Four models generate true editable SVGs from a prompt, not PNGs. When a deliverable needs real vector paths, and how to prompt those models well.
7 min read
The December Deadline Nobody Diarised: Grandfathered Tools and Article 50
A tool you've used since before August 2026 has a separate compliance clock running until December. Here's what changes, and what doesn't wait.
8 min read
The Effort-Visible Edit
A disclosure label alone tends to cost engagement. A study on AI-content disclosure points at the actual lever — and it's a format, not a caveat.
7 min read
The European Accessibility Act and Your Video Library
The Act's main deadline already passed in 2025. A narrower grace period runs to 2030 for unchanged services — how to triage a back catalogue against both.
7 min read
The Parasocial Cost of an AI Tag, and How to Pay It Down
New research traces exactly why an AI disclosure costs engagement — and rules out the two explanations creators usually assume are the cause.
7 min read
The Real Cost of a Reshoot in a Generated Pipeline
A client's small tweak and a client's swap-the-look request aren't the same change order. A credit taxonomy for what each revision type actually costs.
7 min read
The TikTok-As-Search-Engine Claim, Checked Against Its Own Data
The freshest survey behind the TikTok-as-search-engine line both supports it and undercuts it in the same report. What the numbers actually say.
7 min read
The Transcript as a Content Asset
Every talking-head video already contains its own transcript. What it's actually raw material for, and why the text file outlasts the burned-in caption.
7 min read
Three Ways to Find a Generation You Made Weeks Ago
search_my_media, fetch_user_generations, and browse_user_library cover non-overlapping ground — pick wrong and the agent returns nothing, not a redirect.
7 min read