标签：#Video

Amazing : Researchers from ETH Zurich in Switzerland, however, managed to create a new type of pixel that can simultaneously do both. This hypercharged pixel, called a Fourier pixel, can generate and sense arbitrary light fields and tap into a pixel’s full potential for carrying information by manipulating light’s intensity, oscillation phases, and polarization. The team reported its findings in a paper published yesterday in Nature. We are one step closer to 1984 technology: The telescreen received and transmitted simultaneously. Any sound that Winston made, above the level of a very low whisper, would be picked up by it; moreover, so long as he remained within the field of vision which the metal plaque commanded, he could be seen as well as heard. There was of course no way of knowing whether you were being watched at any given moment...

But with great power comes a great price.

With the cash, the company aims to expand its world model offering and reach customers across geographies.

Why async matters for video I've been running useKnockout - a background removal API that processes images in ~200ms - for a few months. Images are fast enough to handle synchronously: POST a file, wait 200ms, get a PNG back. Video is different. Even a 5-second clip at 30fps is 150 frames. At 200ms per frame, that's 30 seconds of processing. You can't hold an HTTP connection open for 30 seconds and call it a good API. So today I shipped POST /video/remove - async video background removal that returns a job ID immediately, processes in the background, and gives you ProRes 4444 (RGB+alpha) when it's done. What shipped As of v0.11.0 (July 10, 2026): POST /video/remove - upload a video, get a job ID back GET /jobs/{job_id} - poll for status, download the result when ready ProRes 4444 output - RGB with full alpha channel, ready to drop into Premiere/Final Cut/DaVinci Node SDK videoRemove() and getJob() in v0.7.0 Python SDK video_remove() and get_job() in v0.7.0 Billing is a dedicated video.seconds meter at $0.10/sec (different from the per-image rate), with a 15-second cap to keep costs predictable. How to use it (Node SDK) import { useKnockout } from ' useknockout-node ' ; import fs from ' fs ' ; const client = useKnockout ({ apiKey : process . env . KNOCKOUT_API_KEY }); // Submit the video const job = await client . videoRemove ({ file : fs . createReadStream ( ' ./input.mp4 ' ) }); console . log ( ' Job ID: ' , job . id ); // Poll until done let status = await client . getJob ( job . id ); while ( status . status === ' processing ' ) { await new Promise ( resolve => setTimeout ( resolve , 2000 )); status = await client . getJob ( job . id ); } if ( status . status === ' completed ' ) { // Download the ProRes 4444 result const video = await fetch ( status . result_url ); const buffer = await video . arrayBuffer (); fs . writeFileSync ( ' ./output.mov ' , Buffer . from ( buffer )); } The job object includes duration_seconds (billed amount), status ( processing / complet

Image-to-video generation is often described as a simple interaction: upload image -> describe motion -> get video That description hides the real problem. A single still contains only one view of a subject. When we ask a model for a fast camera orbit, a full-body walk, or expressive gestures, we are asking it to invent information that was never present in the source. That is where identity drift, unstable lighting, texture flicker, and waxy faces come from. The useful way to approach Seedance 2.0 image-to-video is not as a prompt-writing contest. It is a constraint-management workflow. Give the model a strong identity anchor, request motion that the source image can support, and evaluate one variable at a time. This post explains that workflow in a way that is useful whether you are animating a product render, a character portrait, an approved client still, or a visual asset for a prototype. Note: Model capabilities, pricing, model availability, and input limits change quickly. Check the current documentation and the terms of the platform you use before committing a production workflow. Why image-to-video is different from text-to-video Text-to-video is excellent when invention is the point. You describe a scene and let the model make creative decisions about characters, lighting, composition, and motion. Image-to-video is the better tool when those decisions have already been made and must remain stable. Situation Better starting mode Why Product hero shot Image-to-video Label, shape, material, and color must remain recognizable Character-led sequence Image-to-video One strong reference can anchor a character across clips Approved campaign still Image-to-video The source already represents the accepted art direction Atmospheric B-roll Text-to-video Exact subject identity matters less than visual exploration Abstract concept film Text-to-video Inventing a scene is more valuable than preserving one Existing brand-photo library Image-to-video Stills become reusable

A surprise 10th-anniversary event saw Niantic fulfilling a promise teased in the original 2016 launch trailer for its popular mobile game.

Provided you have a library of SNES cartridges, the SN Operator is a seamless plug-and-play system for easy ’90s nostalgia.

TL;DR We're building a script that takes a video in English and produces the same video narrated in Spanish, in a cloned version of the original speaker's voice. Stack: faster-whisper for timestamped transcription, an LLM (or any MT engine) for translation, XTTS-v2 for voice-cloned synthesis, FFmpeg for surgery. We'll also handle the problem every demo skips: translated audio that doesn't fit its time slot. 📦 Code: github.com/USER/repo (replace before publishing) If you'd rather start from a finished system, Softcatala's open-dubbing and KrillinAI are full pipelines behind one CLI. This post builds the minimal version by hand so you understand what those tools are doing, and where they break. 0. Setup and a licensing warning ⚠️ Python 3.10–3.12. The original Coqui company shut down in early 2024; the maintained fork of their TTS library is published by Idiap as coqui-tts : $ python -m venv dub && source dub/bin/activate $ pip install faster-whisper coqui-tts $ ffmpeg -version | head -1 # 6.0+ is fine, 8.x current ⚠️ Note: the XTTS-v2 model weights ship under the Coqui Public Model License, which restricts commercial use. Prototype freely, but before dubbed videos ship to paying customers, someone must read that license and possibly swap the synthesis step for a commercially licensed model or paid API. Voice cloning also requires the speaker's consent. Get it in writing. 1. Extract audio and transcribe with word timestamps 🎙️ # pull mono 16k audio for the ASR step $ ffmpeg -i input.mp4 -vn -ac 1 -ar 16000 -y source.wav # dub/transcribe.py from faster_whisper import WhisperModel model = WhisperModel ( " large-v3-turbo " , compute_type = " int8 " ) segments , info = model . transcribe ( " source.wav " , word_timestamps = True ) lines = [] for seg in segments : lines . append ({ " start " : seg . start , " end " : seg . end , " text " : seg . text . strip (), }) print ( f " language= { info . language } segments= { len ( lines ) } " ) The timestamps are the skeleton of

TL;DR FFmpeg 8.x includes av1_vulkan , the first cross-vendor GPU AV1 encoder in mainline FFmpeg. We'll probe whether your GPU + driver actually expose AV1 encode, run a first working encode, benchmark it against SVT-AV1 on your own content, and talk about which jobs deserve it. 📦 Code: github.com/USER/repo (replace before publishing) Until FFmpeg 8.0 ("Huffman", released August 2025), GPU AV1 encoding meant picking a vendor: av1_nvenc for NVIDIA RTX 40+, av1_amf for AMD, av1_qsv for Intel Arc. Three code paths, three sets of flags, three driver stacks. The Vulkan Video encode work gives FFmpeg one encoder that reaches all three vendors through the standard VK_KHR_video_encode_av1 extension. The catch: driver support is a lottery. Plenty of capable hardware sits behind drivers that don't expose the encode extension yet. So before any pipeline decisions, we probe. 1. Check what you're running You want FFmpeg 8.x (8.1.2 is current as of late June 2026) built with Vulkan support, plus the vulkaninfo tool from the Vulkan SDK / vulkan-tools package. $ ffmpeg -version | head -1 ffmpeg version 8.1.2 Copyright ( c ) 2000-2026 the FFmpeg developers $ ffmpeg -hide_banner -encoders | grep vulkan V....D av1_vulkan AV1 ( Vulkan ) ( codec av1 ) If av1_vulkan doesn't appear, your build wasn't compiled with --enable-vulkan (distro packages vary; the BtbN static builds and most 8.x distro packages include it). 2. Probe the driver for AV1 encode 🔍 The encoder existing in FFmpeg means nothing if the driver doesn't expose the extension. This is the step that separates "should work" from "works": $ vulkaninfo | grep -iE "video_encode_(av1|queue)" VK_KHR_video_encode_av1 : extension revision 1 VK_KHR_video_encode_queue : extension revision 12 You see Meaning Both extensions listed You can encode AV1 via Vulkan 🎉 Only video_encode_queue Driver does Vulkan encode, but not AV1 (maybe H.264/H.265 only) Neither Driver too old, or GPU lacks an AV1-capable video engine Rough hardware floor: the

TL;DR We're adding live streaming to a SaaS dashboard: a backend endpoint that creates a stream, OBS as the broadcaster over RTMPS, LL-HLS playback with hls.js, and a webhook handler that keeps the UI honest. Working "go live" flow in an afternoon. 📦 Code: github.com/USER/repo (replace before publishing) Webinars, coaching sessions, company town halls: sooner or later your product gets the "can users go live?" ticket. The hard parts (ingest servers, transcoding, CDN delivery) are exactly the parts you should not build. We'll use FastPix as the managed layer here; the same flow works nearly line-for-line on Mux, Cloudflare Stream, or api.video. What we're building: A backend endpoint that creates a live stream and returns a stream key An OBS setup broadcasters can follow in two minutes A viewer page playing LL-HLS with hls.js A webhook handler that flips the webinar between scheduled → live → ended 1. Create the stream server-side 🛠️ You need API credentials (Access Token ID + Secret Key). FastPix uses Basic auth on the server API. Node 20.x, plain fetch , no SDK required (though official Node.js/Python/Go/Ruby/PHP/Java/C# SDKs exist if you prefer). // server/routes/streams.js import { Router } from " express " ; const router = Router (); const AUTH = " Basic " + Buffer . from ( ` ${ process . env . FP_TOKEN_ID } : ${ process . env . FP_SECRET } ` ). toString ( " base64 " ); router . post ( " /webinars/:id/stream " , async ( req , res ) => { const r = await fetch ( " https://api.fastpix.io/v1/live/streams " , { method : " POST " , headers : { " Content-Type " : " application/json " , Authorization : AUTH }, body : JSON . stringify ({ playbackSettings : { accessPolicy : " public " }, }), }); if ( ! r . ok ) return res . status ( 502 ). json ({ error : " stream create failed " }); const stream = await r . json (); // persist against your webinar row: // streamId, streamKey (SECRET!), playbackId await db . webinar . update ( req . params . id , { streamId : stream . str

Part 1: Solving GOP Structure and Compatibility Issues Operating a digital television network isn't just about keeping channels on air—it's about maintaining quality that viewers expect and troubleshooting issues before they escalate. When something goes wrong in live broadcasting, every second counts. But how do you quickly pinpoint whether the problem lies in encoder settings, transport stream structure, or temporal metadata? This is where specialized stream analysis tools become essential. In this series of articles, we'll walk through real-world scenarios that broadcast engineers face daily and show practical approaches to diagnosing and resolving them. When File Analysis Becomes Critical While live monitoring catches issues as they happen, file-based analysis is your diagnostic microscope. Here's the typical workflow: something breaks in production, engineers capture a few minutes of the problematic stream, and now they need to understand exactly what went wrong. File analyzers serve three primary purposes: Troubleshooting: Identifying the root cause of broadcast issues Encoder optimization: Fine-tuning compression settings Quality control: Validating compliance with standards and specifications Let's explore how this works in practice with actual tools and techniques. The GOP Structure Problem Here's a scenario every broadcast engineer has encountered: legacy set-top boxes or older TV models suddenly can't play your stream. The audio works, video starts and stops, or you see freezing. The culprit? Often, it's the GOP (Group of Pictures) structure. H.264 has been around since 2003—over 20 years. Almost everything supports it, yet you'll still find legacy equipment that struggles with certain configurations. Specifically, the number of B-frames can make or break compatibility. Why B-frames matter: They enable lower bitrates while maintaining quality by increasing encoding complexity through bidirectional prediction. But this comes at a cost—a more complex refere

📦 Code: github.com/USER/hls-multi-audio - replace before publishing TL;DR We'll add a working language picker to an HLS player. The hard part isn't the dropdown, it's the manifest. We'll author alternate audio with EXT-X-MEDIA audio groups, package it correctly, debug the classic "zero audio tracks" bug, and wire a switcher on hls.js v1.7 . Adaptive video, captions, the whole pipeline already works. Now someone wants an English/Spanish audio toggle. In HLS, "which audio can the viewer pick" is decided at packaging time and written into the master playlist. The player just displays it. Let's build it in that order. 1. Understand the structure (audio groups) HLS decouples video variants from audio renditions: Each audio rendition is an #EXT-X-MEDIA:TYPE=AUDIO entry pointing at its own media playlist. Renditions are bundled into a named audio group via GROUP-ID . Each video variant ( #EXT-X-STREAM-INF ) references a group with AUDIO="..." . A correct master playlist: #EXTM3U #EXT-X-VERSION:6 #EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud",NAME="English",LANGUAGE="en",DEFAULT=YES,AUTOSELECT=YES,CHANNELS="2",URI="audio/en.m3u8" #EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud",NAME="Espanol",LANGUAGE="es",DEFAULT=NO,AUTOSELECT=YES,CHANNELS="2",URI="audio/es.m3u8" #EXT-X-STREAM-INF:BANDWIDTH=2128000,CODECS="avc1.640028,mp4a.40.2",AUDIO="aud" video/720p.m3u8 #EXT-X-STREAM-INF:BANDWIDTH=1128000,CODECS="avc1.640020,mp4a.40.2",AUDIO="aud" video/480p.m3u8 Every attribute earns its place: LANGUAGE - BCP-47 code, used for the label. DEFAULT - plays when the viewer has no preference. AUTOSELECT - may be auto-picked from the OS language. CHANNELS - needed so the player can reason about stereo vs surround. BANDWIDTH on each video variant must include the audio group's bitrate , or your ABR logic works from a wrong total. 2. Author the renditions with FFmpeg Extract/encode each language's audio, then package. First, encode video-only and audio-only renditions: # video only (no audio), two ladder rungs

📦 Code: github.com/USER/nvenc-vs-cpu-bench - replace before publishing TL;DR A GPU encodes faster than a CPU, but "faster" and "cheaper" are different claims. We'll build a small FFmpeg + VMAF harness that times software (libx264/SVT-AV1) against hardware (h264_nvenc/av1_nvenc), then plug the results into a dollars-per-encoded-minute formula so you find your break-even instead of trusting a benchmark blog. We're using FFmpeg 7.1.x (current stable line) and an NVIDIA GPU with NVENC. Same approach works for Intel QSV ( *_qsv ) and AMD AMF ( *_amf ) if you swap the encoder names. Why this isn't obvious NVENC is a fixed-function hardware block, not "the GPU doing x264 in parallel." It's extremely fast and barely touches the CPU, but it exposes fewer rate-control knobs and gives up a little compression efficiency versus a slow software preset. The gap has narrowed a lot, but it's still there at the quality-obsessed end. So the decision is per-job, and it comes down to one number: dollars per encoded minute = (instance $/hr) ÷ (minutes encoded/hr) . GPU instances cost more per hour but encode many streams in parallel, so the answer depends on whether you can keep the encoder saturated. Let's measure instead of argue. 1. Set up the encoders Three contenders. One representative source file (use real footage, not a synthetic clip). # software H.264, quality-leaning preset ffmpeg -y -i source.mp4 -c :v libx264 -preset slow -crf 21 -an out_cpu.mp4 # NVENC H.264, quality-tuned ffmpeg -y -hwaccel cuda -i source.mp4 -c :v h264_nvenc -preset p6 -tune hq \ -rc vbr -cq 23 -an out_gpu.mp4 # AV1: software (SVT-AV1) vs hardware (needs Ada / RTX 40+) ffmpeg -y -i source.mp4 -c :v libsvtav1 -preset 6 -crf 30 -an out_svtav1.mp4 ffmpeg -y -hwaccel cuda -i source.mp4 -c :v av1_nvenc -preset p5 -cq 30 -an out_av1nvenc.mp4 💡 Tip: -preset p1 (fastest) through -preset p7 (slowest/highest quality) for NVENC. p6 / p7 is where it competes on quality; p1 - p3 is where it competes on raw throughput.

📦 Code: github.com/USER/video-api-bench - replace before publishing TL;DR The per-minute delivery rate is the easiest number to compare and the least useful. The real cost lives in encoding, analytics, and the player. This post compares Mux, Cloudflare Stream, api.video, FastPix, and AWS on what each includes by default, then gives you a tiny script to benchmark upload and time-to-ready on your own files so you stop trusting marketing pages. I have shipped video on four managed APIs across three jobs, and every single time the invoice surprised someone. Not because the delivery rate was wrong, but because encoding, analytics, and the player turned out to be separate line items on some platforms and free on others. Let's compare the parts that don't show up in the headline number. ⚠️ Note: pricing pages move. Everything here was checked in June 2026; verify the links before quoting numbers. 1. Encoding: free or metered? This is the widest spread in the whole comparison. Platform Encoding Delivery Storage Cloudflare Stream Free $1 / 1,000 min delivered $5 / 1,000 min stored api.video Free (unlimited) $0.0017 / min $0.00285 / min FastPix Free on standard plan ~$0.00096 / min @1080p Per-minute, tiered Mux Metered per minute Per minute Per minute AWS (DIY) Per minute (MediaConvert) Per GB (CloudFront) Per GB (S3) If your catalog is upload-heavy (lots of assets encoded once, watched rarely), metered encoding is not a rounding error. It can flip which platform is cheapest, even when the delivery rates look identical. 2. Analytics: included or a $499 floor? QoE analytics is the feature teams forget to price until playback breaks in production. Platform QoE analytics Entry cost FastPix (Video Data) Session-level, 50+ signals/session Free up to 100K views/month Mux (Mux Data) Mature, broad device SDKs $499/month (Media plan, 1M views, +$0.50/1K) Cloudflare Stream Basic Included, limited depth api.video Available Usage-based AWS Build it yourself (CloudWatch + logs) Engineerin

TL;DR If your product lets strangers upload video, you need moderation before launch, not after the first bad upload. We will build a small-team pipeline: extract frames with FFmpeg, score them with NudeNet (ONNX Runtime, CPU-friendly), route uploads into approve / human-review / block by confidence, and log every decision. No trust-and-safety department required. 📦 Code: github.com/USER/ugc-moderation, replace before publishing ⚠️ Note: this is a sensitive area. The goal here is the engineering shape (sampling, scoring, routing, auditing), not detection of any specific content. Keep test fixtures clean and lawful. A model does not decide what is allowed. It produces a score. You decide where the lines go. The whole design is about routing scores sensibly and sending the uncertain middle to a human. The architecture 🧠 upload ──> extract sample frames (ffmpeg) ──> score frames (NudeNet / ONNX) ──> aggregate to one confidence ──> route: high-confidence clean -> auto-approve uncertain middle band -> human review queue high-confidence violation -> auto-block ──> write an audit record for every decision The economics only work if the middle band is small. A decent model makes most uploads obviously fine or obviously not, so a human only ever sees the genuinely ambiguous slice. 1. Sample frames, do not score every frame You cannot afford every frame and you do not need it. Pull one frame per second (or scene-change keyframes) with FFmpeg. # extract 1 frame per second into ./frames mkdir -p frames ffmpeg -i upload.mp4 -vf "fps=1" -q :v 3 frames/frame_%05d.jpg Prefer scene changes to catch more variety with fewer frames: # keyframes where the scene actually changes ffmpeg -i upload.mp4 -vf "select='gt(scene,0.3)',showinfo" -vsync vfr frames/scene_%05d.jpg 💡 Tip: a 30-minute upload at 1 fps is ~1,800 frames. Scene-change sampling often cuts that by an order of magnitude with little loss for moderation purposes. 2. Score frames with NudeNet NudeNet runs on ONNX Runtime on pla

TL;DR Converting HDR10 to SDR with a naive FFmpeg command gives you grey, washed-out video. The fix is tone mapping. We will detect HDR with ffprobe , run two working tone-map chains ( zscale on CPU, libplacebo on GPU) in FFmpeg 8.0, compare operators, and batch it. Test the commands on your own build before shipping. 📦 Code: github.com/USER/hdr-to-sdr, replace before publishing If you have ever run an HDR clip through your normal pipeline and gotten back something flat and foggy, this post is for you. The bug is that HDR and SDR are different color systems, and "just converting" reinterprets one as the other. We will use FFmpeg 8.0 "Huffman" (8.0.2 is current as of May 2026). Why naive conversion fails HDR10 SDR Transfer function PQ (SMPTE ST 2084) gamma 2.4 / BT.1886 Color primaries Rec.2020 (wide) BT.709 (narrow) Peak luminance ~1,000 to 4,000 nits ~100 nits A command that ends in -pix_fmt yuv420p with no tone mapping reads PQ-encoded, Rec.2020 values as if they were SDR. The gamut gets crushed with no intelligence and the brightness curve is misread. Hence the fog. 1. Detect whether a file is even HDR 🔍 Do not tone-map SDR files. Check first: # detect transfer characteristics and primaries ffprobe -v error -select_streams v:0 \ -show_entries stream = color_transfer,color_primaries,color_space \ -of default = noprint_wrappers = 1 input.mkv HDR10 content reports something like: color_space = bt2020nc color_transfer = smpte2084 color_primaries = bt2020 If color_transfer is smpte2084 (PQ) or arib-std-b67 (HLG), you have HDR and you need to tone-map. If it says bt709 , leave it alone. 2. The libplacebo path (GPU, my default) 🚀 libplacebo is the Vulkan-accelerated filter in FFmpeg 8.0. It follows the ITU tone-mapping recommendations and handles the color conversions internally, so the command is short: ffmpeg -i input.mkv \ -vf "libplacebo=tonemapping=bt.2390:colorspace=bt709:color_primaries=bt709:color_trc=bt709:format=yuv420p" \ -c :v libx264 -crf 20 -c :a copy \ ou

The Problem Nobody Talks About Everyone's posting AI-generated videos — characters speaking with lip-sync, manga panels coming alive, virtual idols dancing. The pitch: "just describe what you want." I tried. For months. Here's what I got: Character's face morphed by frame 2 "Slowly looks up" became "violent head shake" Voice-over sounded like Google Translate Same prompt, 3 runs, 3 completely different results No idea what to include or how long the prompt should be Tutorials were either too vague ("be detailed") or too technical (parameter tuning from line 1). The real issue: video prompts are structurally different from text/image prompts. You need to simultaneously control visuals, motion, audio, camera, and consistency constraints — in the right order, at the right length. What I Found A Skill in the Model Studio official repo called happyhorse-prompt-studio . It doesn't teach you theory — it asks you questions and assembles the prompt for you . 4-phase flow: 1. Inspiration Menu Shows you 4 "flavors" of what HappyHorse can do: Flavor What it does A · Voiced Manga Drama Characters talk to each other, with voice + lip-sync B · Character Voice PV Single character self-introduction, 8-10 sec C · Manga Panel Motion Static manga panel starts breathing D · Virtual Idol MV Idol performance with choreography 2. Discovery Asks you conversationally: character appearance, scene, emotion, dialogue, voice type, art style, camera. 3. Prompt Assembly Assembles using the HappyHorse Formula : Scene + Subject + Motion + Audio + Quality Key techniques: @「Image n」 syntax locks character identity across shots Dialogue ≤15 characters (split shots if longer) Japanese prompts work best (HappyHorse is JP-optimized) Always end with キャラの顔・髪・衣装が変わらない (face/hair/outfit stays unchanged) 4. Quality Check Auto-reviews: completeness, compliance, cost estimate, optimization tips. Before vs. After Dimension Writing myself With Prompt Studio Attempts needed 10-20 before one usable 2-3 to satisfacti

Google added two new models to the Gemini API today: Nano Banana 2 Lite (image generation) and Gemini Omni Flash (video generation + editing). Neither is the Gemini 3.5 Pro release people have been waiting for, so it's easy to miss. Here's what's actually in them. TL;DR Nano Banana 2 Lite: gemini-3.1-flash-lite-image = text-to-image in ~4s, $0.034/1K images Gemini Omni Flash: gemini-omni-flash-preview = video gen + conversational editing, $0.10/sec Both are built to be chained: generate an image fast, then animate it into video Neither model is positioned as a quality upgrade = both are cost/speed plays Nano Banana 2 Lite Model ID: gemini-3.1-flash-lite-image Text-to-image output in about 4 seconds $0.034 per 1K-resolution image Positioned as the direct replacement for the original Nano Banana ( gemini-2.5-flash-image ) - if you're on that model, this is a drop-in upgrade Available in Google AI Studio, Gemini API, Gemini Enterprise Agent Platform, and consumer surfaces (Search AI Mode, Gemini app, Photos, NotebookLM, Flow, Google Ads) Gemini Omni Flash Model ID: gemini-omni-flash-preview Public preview in Google AI Studio and the Gemini API Conversational editing - refine a generated video using plain-language instructions instead of re-prompting from zero Multimodal referencing - combine text, image, and video inputs to keep a scene consistent $0.10 per second of video output (same rate as Veo 3.1 Fast) Known limitations right now Generations capped at 10 seconds No audio reference uploads yet No scene extension yet Video references under 3 seconds are accepted by the API schema but not correctly processed yet Character consistency across scene changes/pans still has rough edges Google says longer durations are coming. The part worth paying attention to: chaining them Generate an image with Nano Banana 2 Lite (fast, cheap) Pass that image as a reference into Omni Flash Omni Flash animates it into a video Both models are optimized for throughput and cost, not for to

I ripped open $892 worth of Pokémon packs on my phone in under 15 minutes and walked away with 62 cents. My adrenaline rush felt like the future of gambling.

Video AI systems consistently fail to track what happens when the camera looks away: when a scene pans away from an object in motion and returns, current models re-render the object in its original position rather than showing the logical result of off-screen change. Scaling to more parameters makes this failure worse, not better, according to WRBench , a new benchmark that tests what researchers call "world model reliability." The benchmark presents AI video systems with scenes where something happens off-screen — the camera pans away while an object is in motion, or while a light changes, or while an open door should stay open — then pans back to see what the system believes should have happened. A system that genuinely models the world would track what occurred during the off-screen interval. Current systems mostly don't. Key facts What: A new benchmark tests whether video AI systems can track what happens to parts of a scene the camera isn't currently showing. Across 23 models, the answer is mostly no — and making the models larger made the problem worse, not better. When: 2026-06-19 Primary source: read the source (arXiv 2606.20545) The benchmark covers twenty-three different video generation models and nearly ten thousand video clips across six categories of off-screen change, each designed to test a different aspect of world continuity: objects in motion, light sources changing, object states such as open or closed doors, and several others. This gives a comprehensive picture rather than a single narrow test. The most striking finding is the scaling result. The researchers tested one of the more capable video generation systems at two different sizes: a smaller version and one with more than ten times as many parameters. More parameters didn't help. Scaling made the off-screen tracking problem measurably worse. The larger model produced more realistic-looking frames, but it was less accurate about what should have happened to the parts of the scene it wasn't