Mert

Google bugün Gemini’a geçişi kolaylaştıran iki faydalı özellik duyurdu. Kullanıcılar artık ChatGPT, Claude ya da Copilot gibi diğer yapay zeka sistemlerinde kayıt altına alınan kişisel bilgilerini kolaylıkla Gemini’a aktarabiliyor. Kullanılan mevcut sistemin, “Yapay zeka asistanları arasında bağlam aktarmama yardımcı oluyorsun. İşin, geçmiş konuşmalarımızı inceleyip hakkımda bildiklerini özetlemek.” prompt’una verdi cevabı analiz ederek yeni bir bellek oluşturan Gemini, bugün buna ek olarak sohbet geçmişini içe aktarma özelliğini de kazandı. Bu konuda yaptığı açıklamada, “Verilerinizi başka bir yapay zeka sağlayıcısından alarak (en fazla 5 GB) Gemini’a yükleyin.” diyen Google, burada .zip formatında yüklemeyi destekliyor. Anthropic de geçtiğimiz hafta Claude’a geçişi kolaylaştıran bir “bellek” özelliği sunmuştu. “Claude’a geçerken her şeye sıfırdan başlamayın.” denilerek duyurulan özellik, şirketin burada paylaştığı prompt’a diğer yapay zeka servislerinin verdiği yanıt temel alınarak çalışıyor. “Başka bir servise geçiyorum ve verilerimi dışa aktarmam gerekiyor. Hakkımda kaydettiğin her şeyi listele.” şeklinde başlayan prompt’a verilen cevabı analiz eden Claude, kullanıcı için yeni […]

Google, Gemini’a geçişi kolaylaştıran iki yeni özellik duyurdu yazısı ilk olarak Teknoloji haberleri – LOG üzerinde yayınlandı.

Dün gece sürpriz bir adım atan Apple, profesyonellere yönelik Mac Pro modelini resmi olarak emekliye ayırdı. 2023’te çıkan M2 Ultra işlemcili mevcut modeli satıştan kaldıran şirket, artık yeni bir Mac Pro’nun gelmeyeceğini duyurdu. Bu adım sonrasında Apple’ın masaüstü tarafındaki en güçlü Mac modeli Mac Studio olurken, cihazın yıl sonunda M5 Ultra ile güncellenmesi bekleniyor.

Apple, Mac Pro’yu resmi olarak emekliye ayırdı yazısı ilk olarak Teknoloji haberleri – LOG üzerinde yayınlandı.

Most people know Netflix as the place you go to binge a series on a quiet evening. But since 2023, the company has been venturing into territory where there are no second takes — live streaming, at a scale that now reaches tens of millions of households simultaneously.

At Tokyo Video Tech #10 “Continuous”, held on October 22, 2025 at the Netflix Tokyo office, members of Netflix’s live streaming team gave their first talk in Tokyo. They walked the audience through what it actually takes to pull off a live broadcast at Netflix scale: the multi-path signal routing from venue to Broadcast Operations Center, the cloud encoding pipeline designed so that entire regions can fail without viewers noticing, and the culture of relentless rehearsal — including deliberate failure injection — that turns each one-shot event into something the team has already practiced dozens of times.

Report by: Katz Sakai

Launched in 2023, Netflix’s live streaming service has grown — through continuous trial and refinement — into a platform capable of reaching tens of millions of households simultaneously.
By skillfully building on the technical foundation and operational expertise developed through its VOD business, Netflix has created an architecture that balances the immediacy and reliability required for live broadcasting.
Above all, the session conveyed a clear message: success in live streaming depends on constant preparation and disciplined operational practice behind the scenes.

Live Streaming and Netflix

When people think of Netflix, on-demand streaming is usually what comes to mind.
However, the company has also been taking on the challenge of live streaming as a new form of entertainment.
Its first step in this direction was the Chris Rock: Selective Outrage comedy special, streamed live in 2023 — the first live event in Netflix’s history, which drew significant attention worldwide. Since then, Netflix has continued to scale its live streaming efforts.
For example, the boxing event Jake Paul vs. Mike Tyson was streamed live to more than 60 million households around the globe.

From the Venue to the BOC: Signal Aggregation

One of the biggest challenges in live streaming is ensuring that the video signals from the event venue are reliably delivered to the Broadcast Operations Center (BOC).

When Netflix first began its live streaming initiatives, all encoding was handled directly at the venue. However, due to various network and infrastructure constraints, achieving scalability proved difficult. Based on these lessons, Netflix transitioned to a model that minimizes on-site processing and consolidates control within the BOC.

Today, signals are transmitted from the venue to the BOC through multiple independent paths. Satellite links, dedicated lines, IP-based transmission (such as SRT), and bonded cellular connections — each with different technical characteristics — are combined to ensure that even if one path fails, the stream remains uninterrupted.
Furthermore, Netflix conducts technical studies for each transmission method, compiling the resulting insights into recommended parameters and operational guidelines that are shared with vendor partners.

At the BOC, the incoming signals are monitored and processed as needed.
This may include inserting the Netflix watermark, adding slates such as “Starting Soon” before the event or “Thanks for Watching” after it, and in the event of a total signal loss, displaying an emergency slate to maintain a consistent viewing experience.

For large-scale international events such as sports or entertainment broadcasts, the BOC also handles multilingual dubbing and audio mixing — producing simultaneous audio tracks in English, Japanese, German, and other languages to deliver regionally tailored streams to global audiences.

Converting BOC Signals into Delivery Formats in the Cloud

When transmitting video signals from the Broadcast Operations Center (BOC) to the cloud, Netflix places particular emphasis on redundancy and synchronization design to ensure uninterrupted streaming.
In live broadcasting, even a slight interruption can significantly impact the viewer experience, making it essential to balance both availability and consistency across the system.

For transmission from the BOC to the cloud, multiple independent paths — such as dedicated lines, IP-based links, and bonded cellular connections — are used in combination. This design ensures that if one route encounters a failure, others can immediately take over.
Within the cloud, the same incoming signal is processed simultaneously across multiple regions, preventing any single point of failure and ensuring high resilience.

Once the signals reach the cloud, they are first processed by encoders and then passed to live packagers, which convert them into delivery-ready formats.
These packagers leverage Netflix’s extensive knowledge accumulated through years of VOD delivery — specifically, its database of device-level capabilities and limitations. In practice, this means the system already “knows” which codecs, resolutions, and streaming formats are supported on each device, and applies that intelligence directly to live packaging.
As a result, Netflix minimizes compatibility risks and ensures that live streams maintain consistent quality and playback success rates across the vast range of viewing devices.

The packagers are also designed to handle the precise timing and segment synchronization requirements unique to live streaming.
All cloud encoders reference a shared epoch (a common timing source), ensuring that segments generated in different regions remain interchangeable. This enables instant failover between regions — if one region experiences an outage, another can seamlessly take over without disrupting the live stream.

In this way, Netflix’s live streaming architecture — built on the three design pillars of redundancy, time synchronization, and device compatibility — maintains high reliability while delivering a consistent playback experience across diverse devices worldwide.

Principles of Delivery: Compatibility, Scale, and Quality

The three key elements that Netflix prioritizes in live delivery are compatibility, scale, and quality.

Compatibility
The devices used by Netflix viewers are incredibly diverse — from smartphones released many years ago to the latest high-end devices — representing thousands of playback environments. As with VOD, Netflix’s live streaming platform is designed with the guiding principle of being playable on as many devices as possible.
To achieve this, the delivery protocol is unified under an HTTPS-based approach, allowing existing Netflix players to stream live content without modification.
Video is encoded in both the widely compatible AVC (H.264) and the more efficient HEVC (H.265) formats, while support for the next-generation AV1 codec continues to expand. This approach ensures that older devices can still play streams smoothly, while newer ones benefit from higher picture quality — achieving a balance between inclusivity and performance.

Scale
Live streaming can attract tens of millions of concurrent viewers accessing the service simultaneously. To handle this immense load, Netflix leverages its proprietary CDN, Open Connect. More than 18,000 servers deployed across 6,000 locations worldwide receive live segments from the origin and deliver them to viewers from the nearest node.
The player references manifests built on a segment template model, minimizing server queries and allowing efficient handling of massive request volumes.
This scalable architecture enables stable delivery even during global events such as Jake Paul vs. Mike Tyson.

Quality
Just as in VOD, the Netflix player continuously monitors network conditions to select the optimal bitrate in real time. If throughput drops during playback, the system immediately switches to an alternative CDN node or adjusts the video bitrate to prevent buffering. Through these adaptive measures, Netflix ensures that viewers around the world can enjoy a smooth, uninterrupted live streaming experience regardless of their network conditions.

Success in Live Streaming Starts with Preparation

Another crucial factor in live streaming is preparation and operational discipline. Unlike VOD, live streaming is a true one-shot performance — there are no retakes. For that reason, the team treats preparation like daily “training,” continuously running tests and rehearsals to stay ready for any scenario.

Building robust test environments
To simulate real-world live events, Netflix developed an internal system capable of generating virtual live streams. This system allows engineers to instantly launch test broadcasts and verify the entire pipeline — from encoding and packaging to CDN behavior — under production-like conditions. It enables safe experimentation with new features and architectural changes without risking live operations.

Practicing resilience through failure testing
Netflix routinely performs failure injection tests, deliberately introducing issues such as network latency, packet loss, and simulated cloud region outages. This culture of anticipating the unexpected strengthens system robustness, ensuring that even during large-scale global events, live streams remain stable and uninterrupted.

Forecasting and real-time adaptability
Every live event begins with audience size predictions powered by machine learning models, allowing teams to pre-provision sufficient cloud and CDN resources. During and immediately before broadcasts, Netflix continuously analyzes real-time viewership trends. If the system detects audience growth exceeding expectations, capacity adjustments are made on the fly to prevent congestion. By integrating thorough pre-event simulations with agile real-time responses, Netflix achieves stable and consistent live delivery under any conditions.

In Conclusion

What stood out most in this session was how Netflix approaches live streaming — a relatively new domain for the company — without relying on individual know-how. Instead, the team has established clearly documented procedures and specifications to ensure operational consistency and reproducibility. Even the signal setup from the venue to the Broadcast Operations Center (BOC) has been systematized into operational guidelines rather than being treated as a “black magic”, enabling the same level of quality and reliability across different countries and production environments.

Equally impressive was how Netflix has effectively leveraged its long-standing technical assets and operational expertise from VOD streaming. By applying its accumulated knowledge of device compatibility and CDN scaling directly to live workflows, the company delivers live events with the same “Netflix-quality” scalability and reliability that define its on-demand service.

Above all, the session conveyed a powerful message: live success is built on daily preparation. Through continuous testing in virtual live environments, deliberate failure injections, and regular rehearsals, Netflix ensures that each one-time broadcast runs smoothly when it matters most.

For more details on Netflix’s live streaming technology and engineering practices, you can refer to their official Netflix Tech Blog.

• Behind the Streams: Three Years Of Live at Netflix. Part 1.
• Behind the Streams: Building a Reliable Cloud Live Streaming Pipeline for Netflix. Part 2.
• Behind the Streams: Real-Time Recommendations for Live Events Part 3.

Join Us

Tokyo Video Tech is an open community — whether you’re a seasoned video engineer or just curious about the field, you’re welcome to join. Follow us on Luma to stay updated on upcoming events.
→ Join the mettup

27/03/2026
Suryaprakash S – Author
Developer

Today we will see about using of Docker for Github project deployment

• Install required tools
  sudo apt update && sudo apt upgrade -y

• Install Git
  sudo apt install git -y

• Install Node.js (IMPORTANT → Node 20+)
  sudo apt remove nodejs npm -y

curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash –

sudo apt install -y nodejs

• Verify
  node -v
  npm -v

• Clone your GitHub project
  git clone https://github.com/yourusername/your-repo.git
  cd your-repo

• Install dependencies
  rm -rf node_modules package-lock.json
  npm install

• Build project (IMPORTANT)
  npm run build

• Install Docker
  sudo apt install docker.io -y
  sudo systemctl start docker
  sudo systemctl enable docker

• Run website using Docker
  docker run -d -p 80:80
  -v $(pwd)/dist:/usr/share/nginx/html
  –name mysite
  nginx

• Verify container
  docker ps

• Open website
  http:your-ec2-public-ip

If you find any conflicts , check here

• Docker install conflict

sudo apt remove containerd containerd.io -y
sudo apt install docker.io -y

• Docker permission denied
  sudo usermod -aG docker ubuntu
  newgrp docker

• Container not running / site not opening
  docker ps
  Test Locally – curl localhost
  Chekc logs – docker logs mysite

• Website not opening in browser
  Fix: Security Group

Allow:

Type Port Source
HTTP 80 0.0.0.0/0

• FINAL DEPLOYMENT COMMAND
  sudo docker run -d -p 80:80
  -v $(pwd)/dist:/usr/share/nginx/html
  –name mysite
  nginx

A classical music festival and an internet infrastructure company might seem like an unlikely pairing. But for the past several years, IIJ (Internet Initiative Japan) — one of Japan’s first commercial internet service providers, founded in 1992 and now a major force in cloud, networking, and enterprise IT — has been quietly pushing the boundaries of live streaming technology. Not for esports or pop concerts, but for opera, chamber music, and orchestral performances in the heart of Tokyo’s Ueno Park.

At Tokyo Video Tech #10 “Continuous”, held on October 22, 2025 at the Netflix Tokyo office, Fumitaka Watanabe from IIJ pulled back the curtain on what it really takes to deliver these streams: the hundreds of meters of cables laid and removed for every show, the custom subtitle tools built because AI couldn’t match the nuance of opera, and the creative networking solutions required when your venue is two floors underground in a World Heritage site.

Report by: Katz Sakai

“Spring Festival in Tokyo” and IIJ

The Spring Festival in Tokyo is a classical music festival founded in 2005 by maestro Seiji Ozawa and others. Each year, over a period of roughly 40 days from mid-March, around 70 performances are held across 12 to 15 venues throughout Ueno Park. In 2020, the festival was canceled due to the pandemic, but in 2021, IIJ embarked on an unprecedented challenge — live streaming every single concert.

Journey So Far

• 2021: The First Year — “Handcrafted Streaming of Every Performance”
  A conference room inside Tokyo Bunka Kaikan was converted into a temporary streaming base. Video feeds from each venue were aggregated via NDI and fiber connections. Performances were streamed in multiple formats including 4K, high-resolution audio, multi-angle views, and with subtitles.
• 2022: Transition to Remote Production
  A permanent base was established at IIJ’s Iidabashi office, where video from the venues was transmitted via SRT. Encoding and distribution were handled from the office’s stable network environment. IIJ also developed a custom web player that allowed viewers to switch camera angles and check program information in real time.
• 2023: Launch of “IIJ Studio TOKYO”
  In October 2022, IIJ built a dedicated in-house streaming studio — IIJ Studio TOKYO — enabling integrated management of video, audio, and network operations. This dramatically improved the overall stability and quality of the live streams.
• 2024: Returning to the Field with Agile Operations
  As long-term studio occupation proved difficult, the team brought switching equipment back to Ueno venues, producing finished video signals on-site and transmitting them to IIJ Studio TOKYO via SRT, which continued to serve as the encoding hub. Operations were reoptimized for rapid setup and teardown at each performance.

Challenges Faced in On-Site Operations

The Spring Festival in Tokyo presents unique challenges due to its long duration and multi-venue nature, requiring IIJ to tackle numerous operational and technical issues.

One of the most demanding tasks was cable management and setup. Connecting multiple venues scattered across Ueno Park required laying and removing hundreds of meters of cables for every performance. Preventing tripping hazards and cable damage demanded meticulous on-site expertise and coordination.

Staff allocation was another major concern. At least two operators were assigned to each concert to ensure continuous coverage, but this structure also increased personnel costs and highlighted the difficulty of balancing efficiency with budget constraints.

The diversity of venue environments posed additional hurdles. Some locations had unstable power supplies, causing equipment malfunctions that required off-site reinspection. Others lacked permanent network connections, making it necessary to arrange mobile or temporary lines months in advance.

Unstable networks also proved problematic — especially during the spring hanami season, when mobile networks in Ueno Park became congested despite the use of bonded multi-carrier connections.

A further challenge specific to classical music was the precision required for subtitles. In opera performances, the accuracy and timing of subtitles are integral to the artistic experience, and AI-generated captions proved inadequate. IIJ developed an in-house subtitle insertion tool, allowing human translators’ scripts to be synchronized precisely with the performance in real time.

In summary, delivering these streams demanded not only technical reliability but also deep sensitivity to the cultural context of the event. The success of the project lay in IIJ’s ability to provide a “culturally informed live streaming experience” — one that preserved the integrity and atmosphere of live classical performance.

New Challenges in 2025

In 2025, IIJ took its previous initiatives a step further by introducing several new technical and operational challenges.

First, the team leveraged Vipe, a cloud-based playout system from BCNEXXT, enabling operators to switch between program segments, intermissions, and closing screens with a single click based on a preset timetable. This greatly reduced operational workload and improved on-site efficiency, though it also revealed a new challenge — rising costs associated with cloud services.

Next, IIJ collaborated with KORG and NHK Technologies to conduct a live stream using Dolby Atmos, delivering an immersive audio experience that traditional stereo streaming could not achieve. This marked a new milestone in IIJ’s efforts to “reproduce the live atmosphere as it is.”

The team also took on the ambitious task of testing live streaming from the National Museum of Western Art, a World Cultural Heritage site. Due to strict regulations, every piece of equipment required prior approval, along with detailed reports on power usage and personnel. The concert hall’s location — two floors underground — posed further difficulties, as neither wired nor mobile networks were available, and new line installations were prohibited. IIJ overcame this by connecting the museum to the nearby Tokyo Bunka Kaikan across the street via a 60 GHz wireless network, successfully transmitting both video and audio.

Another major constraint was time. Since the museum remained open to the public until 5 p.m., and the concert opened at 6:30 p.m. and began at 7 p.m., setup time was limited to just one and a half hours, demanding exceptional efficiency and coordination.

Finally, to address mobile network congestion during the cherry blossom season, IIJ experimented with local 5G connectivity. The trial confirmed that stable 60 fps video transmission was possible, though it also revealed that operating local 5G in Japan involves complex licensing procedures, equipment preparation, and substantial costs and lead time.

Network Design Philosophy

For the venue network, IIJ implemented redundancy using multiple connections such as FLET’S Hikari Cross and FLET’S Hikari Next, securing two routing paths — IPoE and PPPoE — and managing them with IIJ’s proprietary router, SEIL. In addition, multiple venues around Ueno Park were interconnected in a mesh topology, enabling remote control of cameras and encoders from any location.

The decision to deploy multiple lines stemmed from practical considerations: during the roughly 40-day festival period, scheduled maintenance by network providers could overlap with event days, and relying on a single line would risk operational downtime.

At the same time, IIJ emphasized that redundancy is not simply “the more, the better.” Excessive investment in multi-layered redundancy could harm overall profitability. Redundancy is a safeguard — a means of maintaining peace of mind and operational continuity. The real challenge lies in striking the right balance between cost efficiency and service reliability.

International Collaboration — Partnership with Berlin Phil Media

Since 2016, IIJ has collaborated with Berlin Phil Media, the streaming subsidiary of the Berlin Philharmonic Orchestra, conducting joint experiments on uncompressed audio transmission and high-quality international live relays. Together, the two organizations have pursued the goal of achieving classical music streaming that surpasses traditional broadcast standards, continuing research and technical verification on advanced delivery systems and cross-border integration.

The greatest lesson IIJ has learned from working with Berlin Phil Media is their unwavering “orchestra-first” philosophy. Every technical decision is made with the musicians in mind — for example, the robotic cameras installed in the concert hall operate completely silently so as not to disturb performers’ concentration. This meticulous attention to the artistic environment inspired IIJ to reaffirm the importance of harmonizing technology and artistry in classical music streaming.

In Conclusion

What left the deepest impression from this session was realizing just how much effort and ingenuity lie behind the seemingly effortless beauty of a concert live stream. From laying cables and designing network routes to fine-tuning subtitles, countless individual actions come together to create the seamless environment that allows us, the audience, to enjoy music online.

It was also inspiring to see how the IIJ team never simply repeats the same routine each year, but continually embraces new technological frontiers — such as cloud-based operations, Dolby Atmos, and local 5G. Their pursuit of stability while still pushing the boundaries of innovation truly embodies IIJ’s mission: to evolve the Internet into a genuine social infrastructure.

Join Us

Tokyo Video Tech is an open community — whether you’re a seasoned video engineer or just curious about the field, you’re welcome to join. Follow us on Luma to stay updated on upcoming events.
→ Join the mettup

Mark Gurman tarafından yapılan son açıklamaya göre Apple, 8 Haziran tarihli WWDC26 kapsamında tanıtılacak iOS 27 güncellemesiyle birlikte kullanıcıların Siri ile bağlamak istedikleri yapay zeka sistemini seçmelerine olanak tanıyacak. Siri şu anda yanıtlayamadığı bir soru olursa, bu soruyu OpenAI ile yapılan iş birliği kapsamında ChatGPT’ye iletmeyi öneriyor. iOS 27 ile kişilerin bu süreçte “Gemini” ve “Claude” gibi diğer yapay zeka sistemlerini de kullanabileceği aktarılıyor. iOS 27 ve macOS 27 işletim sistemleriyle birlikte Siri için başka önemli duyurular da yapılacak. Yeni işletim sistemlerine Siri’ye özel ayrı bir uygulama ekleyeceği belirtilen Apple, Google iş birliği kapsamında Gemini tabanlı özel bir yapay zeka modeliyle güçlendireceği Siri’yi tam anlamıyla bir sohbet botu olarak ChatGPT ve Gemini’ın karşısına da konumlandıracak. Geliştirilen bu altyapıda Siri’yle hem sesli hem de yazılı görüşmeler gerçekleştirilebilecek, sorulan sorulara cevaplar alınabilecek. Siri’yi doğrudan Dynamic Island’a konumlandırmayı düşünen Apple, iddiaya göre WWDC26’da asistanın yeni arayüzünü görücüye çıkaracak ve yine Gurman’a göre Siri’ye […]

Apple, tüm ana yapay zeka sistemlerinin Siri’ye entegre edilmesine izin verebilir yazısı ilk olarak Teknoloji haberleri – LOG üzerinde yayınlandı.