Prasar Bharati opens doors to AI startups for WAVES, Doordarshan and Akashvani content

Prasar Bharati is looking to build a technology-driven content ecosystem around artificial intelligence, proposing to empanel agencies, studios, startups and technology companies for AI-enabled content creation, multilingual localisation, archive repurposing, restoration and digital-human technologies across its broadcast, OTT and digital platforms.

The public broadcaster has issued a proposed Expression of Interest (EOI) for empanelment of organisations that can support its evolving content and technology requirements across WAVES, Doordarshan, Akashvani, DD India and other platforms.

The proposed framework is explicitly designed to be startup-friendly, with Prasar Bharati stating that it intends to prioritise demonstrated capability over traditional financial thresholds. Importantly, the EOI does not prescribe any minimum turnover requirement. Financial and commercial eligibility requirements, where necessary, would be specified separately for individual assignments or subsequent RFPs.

The move comes as Prasar Bharati expands its digital ecosystem, including WAVES OTT, launched in November 2024, alongside digital publishing, archive modernisation and multilingual content initiatives. The broadcaster said the rapid evolution of AI is creating opportunities to accelerate production, expand multilingual reach, improve accessibility and unlock the value of its decades-old broadcast archives.

Nine categories for AI-enabled content ecosystem

Under the proposed EOI, applicants can seek empanelment across nine categories, with evaluation to be conducted independently for each category.

These include short-form content of 30 seconds to three minutes; one-to-five-minute micro dramas; two-to-six-minute music videos; eight-to-10-minute intermediate series; 25-to-30-minute regular series; feature films of 90 minutes or more; archive repurposing and restoration; AI localisation, translation and dubbing; and AI avatars, digital humans and virtual presenters.

The scope goes beyond using AI simply as a production tool. Prasar Bharati is proposing to explore AI-generated video, audio-to-video and image-to-video workflows, AI presenters, interactive and personalised media, AI-native original intellectual property and technology-led audience engagement.

The broadcaster also wants AI to be used to make its archive more accessible. Its Doordarshan and Akashvani archives span several decades of cultural, historical and journalistic material. The proposed services include denoising, colour enhancement, upscaling, reframing, metadata tagging and contextual re-presentation of legacy assets for contemporary audiences and digital platforms.

Digital humans, AI anchors come with stricter safeguards

One of the more significant elements of the proposed framework is the creation of a dedicated category for AI avatars, digital humans and virtual presenters.

Companies seeking empanelment in this category would need to demonstrate a consent framework for any digital human or avatar based on the likeness, voice or persona of a real individual. They would also have to establish a policy for synthetic newscasters and AI presenters, including disclosure to audiences that the presenter is AI-generated.

Applicants would further need the technical ability to deliver multilingual digital-human solutions with real-time or near-real-time lip synchronisation across at least three Indian languages, besides demonstrating at least one previous deployment of an AI anchor or digital human in a broadcast, OTT or digital publishing environment.

The proposed safeguards are particularly stringent for deepfakes and synthetic media. Applicants would have to commit not to produce deepfake content depicting living public figures, elected officials or religious figures for Prasar Bharati without explicit informed consent and prior written approval from the broadcaster.

Every AI-generated or AI-assisted deliverable would also have to carry a machine-readable provenance record, such as C2PA or an equivalent standard, identifying the AI tools, models and workflows used. Human editorial sign-off would be mandatory before AI-generated content is considered broadcast-ready.

For AI anchors and synthetic presenters, Prasar Bharati proposes mandatory on-screen disclosure that the presenter is AI-generated.

AI sovereignty gets a place in the framework

The EOI also signals a push towards greater technological sovereignty in the public broadcaster's AI ecosystem.

Applicants would have to disclose the primary AI models, platforms and APIs used in their production workflows, including their country of origin and licence type. Companies would also have to provide contingency plans explaining how content could be maintained or reproduced if an AI model is discontinued, its licensing conditions change or access is restricted.

Applicants relying exclusively on foreign cloud platforms would be required to explain the implications for data residency and sovereignty.

While use of Indian-developed AI is not made a mandatory eligibility condition, Prasar Bharati says preference may be given to companies using Indian-developed or open-source AI technologies, India-hosted deployments and architectures that reduce dependence on foreign model providers. The framework proposes to award marks for AI sovereignty as part of its technical evaluation.

No turnover barrier for startups

The EOI is designed to widen participation beyond established production houses.

Eligible applicants include companies, LLPs, partnership firms, proprietary firms, DPIIT-recognised startups, studios, creative agencies, technology companies, post-production houses, animation and VFX firms, digital-human and avatar studios, localisation and translation companies and consortiums led by eligible entities.

Prasar Bharati says the evaluation will follow a "capability-first, turnover-agnostic" philosophy, with emphasis on output quality, workflow maturity, legal readiness, team competence and delivery capability rather than legacy credentials.

The proposed framework also includes a paid pilot track for emerging companies that may not have previous broadcast-scale delivery experience. Shortlisted applicants could be asked to produce a 60-to-120-second original clip, digital-human sample or localisation sample against a Prasar Bharati brief. A satisfactory pilot could qualify an applicant for provisional empanelment.

Technical capability carries highest weight

Applications would be evaluated out of 100 marks, with a minimum overall score of 70 required for empanelment.

Technical capability carries the highest weight at 40 marks, followed by creative and editorial capability at 20, execution and delivery at 15, pitch or demonstration at 15, and institutional and compliance profile at 10.

Applicants would have to secure at least 25 of the 40 technical marks, 12 of 20 creative and editorial marks and eight of 15 execution and delivery marks.

The technical assessment covers areas including text-to-video, audio-to-video and image-to-video generation, character consistency, multilingual lip synchronisation, multi-character scene management, broadcast-safe mastering, OTT delivery, localisation, AI-anchor capability, provenance controls and AI-model transparency.

The broadcaster has also proposed assessing whether companies use Indian-developed or India-hosted AI systems and whether they have adequate data residency safeguards.

Archive restoration could become a major AI opportunity

The proposed framework gives archive restoration its own evaluation structure, recognising the scale and sensitivity of Prasar Bharati's legacy content.

Applicants under the archive category would be evaluated separately on a 100-point framework. Technical restoration capability would account for 30 marks, metadata and cataloguing 20, rights and clearance 15, archive experience and preservation standards 20, and demonstration through a restoration sample 15.

The sample would involve restoration and re-presentation of 60 to 120 seconds of legacy SD or deteriorated material. A minimum score of 70 would be required for empanelment in the category.

The services envisaged include AI-assisted denoising, colour grading, SD-to-HD and HD-to-4K upscaling, grain management, aspect-ratio conversion and metadata extraction using standards such as PBCore, Dublin Core or equivalent.

This could potentially allow content sitting in legacy Doordarshan and Akashvani archives to be repackaged for OTT consumption, with the EOI specifically referring to structured metadata required for ingestion into WAVES.

Multilingual push across Indian languages

Language localisation is another central pillar of the proposed ecosystem.

Prasar Bharati wants agencies capable of translation across Indian and international languages, including scheduled and tribal languages. The scope includes dubbing, voice adaptation, voice cloning subject to consent, lip synchronisation, captioning, subtitles and accessibility services.

The EOI specifically calls for capabilities covering regional language authenticity and dialectal adaptation, suggesting that the public broadcaster sees AI-assisted localisation as a way to significantly expand the reach of its content across India's linguistic markets.

Rights, data and AI governance built into procurement

The proposed empanelment framework places substantial emphasis on rights management and data security.

Applicants would need to demonstrate copyright governance, music rights management, AI-model licensing compliance, training-data governance and third-party rights clearance. They would also need to document the provenance of AI-generated material.

On the data-security side, agencies would have to maintain secure storage, access controls, encryption, audit trails and incident-response mechanisms. Any security breach involving Prasar Bharati material would have to be reported immediately, with a root-cause and remediation report within 72 hours.

The proposed framework also prohibits storage or processing of Prasar Bharati archive material outside India without explicit written consent, except where technically required by an AI workflow and disclosed upfront. Companies would additionally have to disclose third-party AI model providers and cloud platforms used in production.

Commercial opportunities will follow separate process

The EOI itself will not determine commercial rates or guarantee business to empanelled companies.

Once technical empanelment is completed, Prasar Bharati may conduct limited RFPs among empanelled agencies for commercial rate discovery. These could involve schedule-of-rates, output-based, milestone-based or deliverable-based pricing.

The broadcaster proposes using a Quality and Cost Based Selection process to establish category-wise benchmark rates. It may also allow other empanelled agencies to match the benchmark rates and join the commercial panel. Rates may subsequently be refreshed, ordinarily every six months, reflecting the rapid changes in AI technology and cost structures.

Assignments could subsequently be distributed based on factors including specialist expertise, language capability, creative suitability, technical capability, turnaround time, capacity, past performance and workload.

Prasar Bharati could also use alternative commercial models, including revenue sharing, royalties, co-production, licensing, original-IP development, pilots, innovation challenges and technology demonstrations.

Prasar Bharati to retain IP in commissioned content

The proposed EOI also sets out a default intellectual-property framework.

For fully commissioned content, Prasar Bharati's default position would be full assignment of copyright in the final deliverable across platforms, territories and formats for the full copyright period.

For co-produced or revenue-share projects, IP could be jointly held or licensed according to financial contribution, while Prasar Bharati would retain at least a perpetual, royalty-free licence for broadcast, digital exploitation and archival purposes.

For archive restoration, restored deliverables and associated outputs would vest in Prasar Bharati, while the restoration methodologies, workflows, software tools and proprietary technologies of the applicant would remain with the applicant. Pre-existing AI tools, proprietary software, model weights and methodologies would similarly remain the agency's property.

Empanelment proposed for two years

Qualifying applicants would be empanelled category-wise, with Prasar Bharati potentially creating separate panels for production and content studios, startups and emerging innovators, archive restoration, music and culture, OTT technology, localisation, digital humans and AI technology partners.

The proposed empanelment would ordinarily remain valid for two years and could be extended depending on performance, continuing requirements, integrity status, technological relevance and operational needs.

Prasar Bharati would conduct periodic performance reviews covering quality, timelines, data security, IP and AI-ethics compliance, responsiveness and technological capability. Agencies could be de-empanelled for material misrepresentation, persistent non-performance, AI-ethics breaches, deepfake violations or other legal and contractual violations.

The proposed framework also makes clear that empanelment itself will not guarantee any work, minimum business or financial commitment from Prasar Bharati. Specific assignments would be taken up through subsequent procurement or commissioning mechanisms.

The EOI therefore represents a significant shift in how India's public broadcaster is looking to source content and technology, combining traditional production capabilities with AI-native startups, localisation specialists, archive-tech companies and digital-human providers.

For Prasar Bharati, the proposed model could create a larger pool of specialised technology partners while allowing WAVES and its wider broadcast ecosystem to experiment with AI-generated formats, multilingual programming and the digitisation and repurposing of its vast legacy archive. 

At the same time, the extensive requirements around consent, provenance, human editorial review, data residency and AI-model transparency indicate that the broadcaster is seeking to balance rapid adoption of generative AI with the heightened editorial, legal and reputational responsibilities of a public-service broadcaster.