2026 Accelerator Project

Objectives:

Champions: BFBS, IET, AP (Associated Press)

Participants: Tesla Technologies, Deviqon Labs, InReality, Norsk

Live contribution from the field is most likely to fail in exactly the situations where it matters most. When a contribution link becomes severely congested, suffers interference or is jammed, video quality can deteriorate rapidly, or the feed can be lost altogether. Yet conventional transport protocols can only respond once that degradation is already under way. At the same time, the trustworthiness of live video is under growing pressure. Real-time deepfakes can now be injected into a stream, and encrypted material intercepted today could be decrypted by quantum computers in future, putting organisational reputation, journalists and their sources at risk. 

Q-Stream Alpha set out to address these challenges together. Its objectives were to anticipate network degradation rather than react to it, to preserve the most important parts of the picture when bandwidth collapses, and to demonstrate, frame by frame, that a live feed is authentic and has not been altered between source and destination. 

Innovation:

Encoding, transport and security are usually developed and deployed as separate layers, and transport protocols such as SRT respond only after a link has begun to fail, with retransmissions adding to the congestion. Q-Stream Alpha treats resilience and trust as a single problem, and its innovation lies in how each layer of the workflow informs the next. A forecast of network conditions gives the encoder time to adapt. The adapted picture is then signed with content credentials, so the signature covers exactly the video that is delivered rather than a master that nobody sees. 

Bringing these layers together led to what the team believes were two world-first demonstrations: live contribution with post-quantum encryption, and live contribution with dynamic C2PA credentials and live detection of missing or corrupted C2PA provenance. Because the approach is codec-agnostic, it can be applied across existing and future contribution workflows rather than being tied to a single encoding format. 

The Solution:

Q-Stream Alpha is built on three pillars: 

The Brain predicts. Using analytics from public or private 5G, a machine-learning model forecasts drops in bandwidth up to two seconds before they occur and passes that forecast to the encoder, which can then adjust in good time rather than after quality has already collapsed. 

The Eye adapts. With that advance warning, AI analysis of the picture determines what matters most in the shot, such as a presenter, and keeps it clear while compressing the background heavily, so the subject remains recognisable even as the video bitrate falls towards 80 kbps. 

The Shield protects. As the live video is produced, its C2PA provenance is continuously generated and updated, allowing the receiving end to verify the source and integrity of the feed frame by frame. Manipulated frames fail verification and are refused rather than displayed. Post-quantum cryptography also addresses the “harvest now, decrypt later” threat, in which encrypted material captured today could become readable as quantum computing develops. 

The complete workflow can run on a single laptop at 30 frames per second, showing that these capabilities can be deployed at the edge without dedicated infrastructure. 

Collaboration: 

Working together, the champions developed the project to address both broadcast and defence requirements. AP brought extensive newsgathering experience, helping the team explore how to maintain live feeds from locations with unreliable connectivity, while BFBS added a defence perspective on what acceptable degradation means when accurate information matters more than picture quality. The IET, in line with its vision to engineer a better world, helped validate the use cases for the wider media industry while maintaining a vendor-agnostic, standards-based approach. It also brought in experts from industry bodies focused on privacy and provenance, helping the team navigate these complex areas. 

The four participants integrated their technologies through open standards and interfaces rather than a proprietary end-to-end stack. These included C2PA for provenance signed with FIPS 204 (ML-DSA) post-quantum signatures, Media over QUIC (MOQ) from encoder to decoder, a 3GPP policy interface to the 5G core secured with mutual TLS, ONNX for the AI models and a shared MQTT control bus. This allowed four participants to work within one workflow and to test it end to end under realistic conditions, and it leaves open integration points that other vendors can adopt, so the approach can grow beyond the companies that built it. 

Proof of Concept 

Q-Stream Alpha was first tested live at AP's London offices, with BFBS reporter Sian Grzeszczyk-Melbourne presenting on a live camera feed and describing her experience of live reporting from hostile environments. A single laptop ran the full workflow, from semantic analysis, 5G network simulation and prediction, through adaptive encoding, C2PA signing and encryption, to verification and playback. 

At IBC2026, visitors could interactively see how the layers work together across three modes using live video and delivery to a mobile phone. In Normal mode the feed is verified frame by frame. In Survival mode bandwidth collapses and the background degrades, while the reporter remains clear and the feed remains verified. In Tamper mode a manipulated frame is detected immediately and refused. 

Network conditions were simulated at both demonstrations based on replaying real 5G analytics data. The prediction and the requests to the 5G core were real, and a 5G core acting on those requests to prioritise the video was shown separately in the lab. 

View the Project's PoC Presentation

Key Learnings 

The project showed that the value of each layer increases when it is designed alongside the others. Anticipating a drop in bandwidth gives the system time to protect the most important content before quality collapses, rather than recovering afterwards, and it is the semantic analysis that tells the system which content to protect. 

Building trust into a live, low-bandwidth workflow brought its own challenge. Post-quantum signatures are large, and initially they consumed two-thirds of the available link. By redesigning how provenance was carried, the team reduced its share of the link from 67% to 26%, allowing the picture bitrate to increase from 79 to 291 kbps on less total bandwidth. This made it possible to combine strong provenance with low-bitrate resilience, rather than trading one against the other. 

Running the whole workflow on a single laptop showed that these capabilities do not depend on dedicated infrastructure and can be deployed at the edge, where they are most needed, putting them within reach of smaller newsrooms and organisations with limited budgets.

Industry Impact 

For news organisations, the value of Q-Stream Alpha lies first in getting on air from locations where connectivity is congested or disrupted and a conventional link would fail entirely. It also protects the organisation's reputation, because manipulated footage is detected and dropped before it can be broadcast. And, since the feed is protected against future decryption, material intercepted today cannot later be read and used against the organisation, its sources or its staff. In sport, the same capabilities support contribution from venues with unreliable 5G, including lower-tier competitions without dedicated infrastructure, and meet rights holders' requirements for robust encryption. 

The same approach applies well beyond broadcasting. Defence and emergency services need critical video over fragile links, while companies making market-sensitive or legally mandated financial disclosures need to establish that a live announcement is authentic and comes from the right source. Because C2PA is an open standard, verification need not be tied to a single vendor’s proprietary ecosystem. 

The project is already building momentum beyond IBC. Less than a month after the showcase, Q-Stream Alpha has been shortlisted for a further innovation programme and an industry award. 

Looking Ahead 

Q-Stream Alpha set out to answer a question that matters to every newsroom, broadcaster and defence organisation working in the field: how do you keep a trusted live feed on air when the network is degraded? The answer brings together prediction, intelligent compression and post-quantum protection into a singular workflow. Because it all runs on a single laptop, it does not depend on the cloud and can be deployed wherever it is needed.  

For decades, live broadcasting has been judged largely by how good the picture looks. As networks become more contested and convincing fakes become easier to make, two further questions matter just as much: did the feed get through, and can it be believed? Q-Stream Alpha’s world-first demonstrations showed that both can be answered together, live and at the edge. As the project moves beyond proof of concept, it points towards a future in which getting the picture through and knowing it can be trusted are no longer separate challenges, but fundamental requirements of live media.