News · ai music — By WATMM Desk —

India's oldest label is using AI to make the music videos it never owned

Saregama owns the songs and recordings for much of Hindi cinema's 1960s and 1970s music, but not the film clips that went with them. It now uses generative AI to make new videos for those songs, and to rework some for younger listeners, while it still argues against unlicensed AI. The label hasn't said whether the original singers' or composers' estates get any say.

Key facts

  • What happened: on Saregama's 4 Aug 2026 earnings call, managing director Vikram Mehra said the label uses generative AI to make music videos for old film songs whose original videos it never owned, at a cost he put as low as INR 70,000 (about 740 US dollars) each.
  • How: a content team uses third-party generative AI tools; one version keeps the original audio with a newer-looking video for Gen X, and another changes the instrumentation but keeps the composition for Gen Z.
  • Why it matters: the label that joined the push against OpenAI in the Delhi High Court is now one of the clearest examples of AI used on a catalogue its owner controls, and nothing it has said covers consent or pay for the original performers.

Saregama, India's oldest record label (founded around 1901), has a gap in its catalogue that no newer rival has: it owns the songs but not the pictures. On the label's earnings call on 4 Aug 2026, managing director Vikram Mehra said Saregama holds "all the rights connected to the song, except the right of the original music video" for many film songs from the 1960s and 1970s, according to Music Business Worldwide. Those clips stayed with the films' owners because, in Mehra's words, there "was no market for it" when the films were made.

His fix is generative AI. Reshooting videos for old songs never paid for itself. Mehra said an AI-made video can now cost as little as INR 70,000, around $740, which "may start making financial sense".

What is Saregama actually making?

Mehra described two versions, each aimed at a different audience, per MBW:

  • For Gen X: the original audio, with a "more contemporary looking" video.
  • For Gen Z: the audio is changed, with a "change in the instrumentation" while keeping the composition, and a new video on top.

The same content team is also making podcast-style audio built around Saregama songs. Mehra's example was a 50-year-old looking back on their life, scored with songs about romance, separation or friendship, a format Saregama plans to license to podcast companies. An ICICI Securities note on the quarter confirms the setup: two internal AI teams, one using third-party generative AI tools to produce "modern music videos and thematic podcasts using library tracks", and one working on internal processes. The note adds that extra spending on AI remains "very low". Neither source names the tools, or any specific song that has been given an AI video.

The scale is large. MBW puts the catalogue at about 180,000 songs and the flagship Saregama Music YouTube channel at roughly 64 million subscribers. Mehra also said older songs are turning up on Instagram in surprising volume, so the label sees demand for a catalogue that has no visuals of its own.

Isn't Saregama fighting AI companies?

It is, and it says the two positions fit together. Mehra told analysts the label supports "licensed innovation" and will keep fighting "unlicensed exploitation" of its music. In May, MBW reports, he called on streaming services to give "no value" to AI-generated music.

In February 2025, Saregama, T-Series and the Indian Music Industry (IMI) trade body asked the Delhi High Court to let them join ANI's copyright suit against OpenAI, Business Today reported. Justice Amit Bansal was wary of widening the case, telling them they could file their own suit. Rolling Stone India summarised the labels' complaint: OpenAI used lyrics, compositions and recordings without a licence, leaving artists, composers and producers without credit or payment. OpenAI says it uses publicly available data and follows copyright law.

That front has got harder. On 24 Jul 2026 the same court refused ANI an interim injunction, holding at a preliminary stage that storing ANI's articles to train ChatGPT falls under the Copyright Act's fair-dealing exception, SpicyIP reports. The ruling concerned news text, the full trial is still ahead, and the reports we read don't say what the court made of the music labels' arguments.

Who gets paid when a 1970s song gets an AI video?

Nobody has said. Saregama's comments cover what it owns, what the work costs and who the audience is. Neither MBW nor the ICICI note mentions whether the singers, composers, lyricists or their estates were consulted, or whether any of them earn more when a song gets a new AI video. Neither says whether the AI videos show anyone who looks like the original actors, or whether the Gen Z instrumentation is played by musicians or generated.

These are fair questions to put to Saregama, because its own case against OpenAI rests on artists going uncredited and unpaid. Saregama is within its rights to use recordings it owns. Whether it is also living up to its own argument depends on answers it hasn't given.

What we think

This is the licensed side of AI, and it still shuts people out. Saregama is drawing its line by ownership, not by technology: AI is fine when the rights holder points it at its own catalogue. That's a coherent position, and more honest than scraping. But on an old film song, ownership and authorship sit with different people. The label holds the rights. The voice, the tune and the words came from people who signed contracts decades before anyone imagined a new video being generated for a few hundred dollars. "Licensed" here means the label licensed itself.

Cheap videos won't mean every song gets one. By our arithmetic, INR 70,000 a video across about 180,000 songs would come to roughly INR 1,260 crore. That is more than three times the INR 300 to 350 crore content budget Mehra gave for the year, and he said AI spending sits inside that budget. So Saregama will choose. Mehra also said the label has used predictive AI for four years to decide what music to acquire, and it would be natural to use the same kind of model to pick which old songs get a new video. Songs that get one gain a second life on YouTube and Instagram. The rest risk fading further. The algorithm that decides, and the data behind it, belong to the label, not the artists.

The Gen Z versions deserve the most scrutiny. A new video around an untouched recording mostly repackages it. Changing the instrumentation creates a new recording of the composition, and the musicians on the original track drop out of it. If session players are re-hired, that's work. If the new parts are generated, an old performance has been swapped for a machine's, and listeners won't be able to tell which happened unless the label labels it.

What would make this fair: say which videos and arrangements are AI-made, in the video description and in the credits; tell surviving artists and estates before a reworked version goes out; and share part of any new income from these versions with the authors and performers, not only pay what an old contract requires. Saregama has asked courts and streaming services to take creators' consent and pay seriously. It can show what that looks like on its own catalogue.

Subscribe to We are the music makers

One email a week: the music-tech and AI-in-music news that mattered, and what we think about it.

Sign-ups open soon.