Crosspoint and Amplify to manage metadata of the RTVE Archive

RTVE has awarded Point de croisement the contract to implement AI capabilities for its archive metadata. This project will draw on the expertise of Multimédia, the company’s digital arm, and Amplify technology.

Since 2020, the RTVE team has been committed to adding automatic metadata to its Documentary Collection due to an increase in programmes based on archive content, such as Cachitos de Hierro y Cromo, Viaje al Centro de la Tele et Conexión Vintage, as well as the implementation of video-on-demand platforms like Lecture RTVE. Following the completion of the previous service contract, the public broadcaster launched a new tender for the management of the archive over the coming years. This service is of vital importance given that the corporation has set itself the medium-term goal of cataloguing 198,220 hours of material from its regional centres.

In collaboration with Amplify, Crosspoint (part of the ES Media group) has been awarded the tender to manage and provide the human and technological resources for the automatic metadata service for audiovisual files in the RTVE Archive. The service that Crosspoint and Amplify will deploy will automatically analyse 20,000 hours of content, including: 18,000 hours of Spanish audio, 5,000 hours of image content and 5,000 hours of multilingual audio. Amplify Seiri, an AI-based solution that automates and improves the management of audiovisual archives, will be used to tackle this project, allowing content to be located and reused in seconds.

Audio, image, and multilingual audio analysis

The technology service, which will be fully deployed in September and launched on 1 October, covers a wide range of audio processing tasks. In addition to audio-to-text transcription, the Crosspoint and Amplify tools will handle the punctuation and capitalisation of transcribed text, as well as named entity recognition (NER), keyword extraction to identify topics covered according to RTVE criteria, and the automatic classification of content, including its type (interview, documentary, news, etc.). They will also provide automatic summarisation based on RTVE requirements, identify the gender of speakers, and recognise the identity of public figures based on their voices. They will also perform semantic segmentation as a form of automatic content transcription by news item for the analysis of news programmes and magazines and by speaker.

From an image analysis perspective, the Crosspoint and Amplify solution recognises people and public figures in scenes, extracts alphanumeric characters (OCR) from images, identifies logos and brands, detects relevant objects, identifies scene types and describes images. Of particular relevance is the analysis of news programmes and magazines, given that the RTVE Archive has requested the identification and detection of the start and end of sections, using its time codes, as well as the identification of visual elements such as the header, headlines, sports section, weather section, credits, live signal, correspondent introductions and copy text at the end of the broadcast.

For multilingual audio analysis, the system transcribes and translates languages spoken in the Balearic Islands, the Basque Country, the Valencian Community, Galicia and Catalonia into Spanish, whether the pieces are in one of these languages or are bilingual. Additionally, the system will be used as a proof of concept (PoC) for specific audio and video analysis, including the detection of effects such as applause, cheering or shouting, violent content and hate speech, image search, and the detection and recognition of musical works.

Crosspoint will be responsible for training RTVE staff to manage this tool independently. They will have access to a support team available for consultations for 12 hours a day, with a maximum response time of 48 hours.

Crosspoint/Crossmedia and Amplify: a strategic alliance

The collaboration between Crosspoint and Amplify on the RTVE Archive project is based on a agreement that will extend the capabilities of metadata and automatic summaries to all types of project within the broadcast and OTT ecosystems in all markets in which both companies operate. As with the Spanish public radio and television project, solutions such as Seiri — a ‘digital documentalist’ capable of automatically analysing audiovisual content and extracting large amounts of structured information through accurate transcription — and GeNews — a generative, AI-based solution capable of generating an audiovisual piece including a script, image selection, basic editing and synthesised voice — will be deployed.

Víctor García, managing director of Point de croisement, highly values the partnership with Amplify: “We are thrilled to have signed a framework collaboration agreement with AMPLIFY — a company led by top-tier professionals we’ve known for decades. They’ve brought their deep expertise into developing highly advanced tools, enabling the execution of complex projects such as the automatic metadata enrichment of the RTVE Archive.

We look forward to continuing this partnership and expanding into new markets, supported by our global presence as part of ES Media Group. In this context, Memnon —also part of the group— plays a strategic role in large-scale archive digitization and the implementation of AI-driven solutions.”

On behalf of Amplify, Aaron López, founder and CEO, adds: “Partnering with a trusted, high-caliber integrator like Crosspoint and Crossmedia enables us to deliver on our most ambitious goals. This collaboration with RTVE is a major milestone — and just the beginning of our shared roadmap to serve public and private media organizations with cutting-edge AI solutions.”