Eloelo Dolphin AI Debuts Multi-Shot Agentic Video Studio

A micro-drama company becomes an AI studio

Eloelo Group is the Bengaluru company behind the vertical drama app Story TV. It has launched Dolphin AI, an "agentic" studio that takes video from script and shot planning through generation and editing in one tool.34 Founder and CEO Saurabh Pandey announced it in his keynote at FICCI FRAMES 2026 in Mumbai on September 29. Prithul Kumar, a Joint Secretary at the Ministry of Information and Broadcasting, opened the company's booth at the event.26 The product is now in early access. People can join a waitlist at trydolphin.ai or use an invite code.24

The news trickled out over more than a week. Trade sites wrote it up on launch day, national business outlets followed in early October, and Mint ran its story on October 8.6101 All of this coverage leans heavily on the company's own framing. No outlet published independent tests, pricing, or output samples. Readers should treat the strongest claims as promises the company has made, not results anyone has checked.

Even so, the launch is worth a close look. It is a clear example of where AI video seems to be heading in 2026: away from single-model demos and toward orchestration software that lets the generated clips be put to real use.

The problem Dolphin targets: continuity

Every outlet describes the same core pitch. Today's text-to-video systems can make a striking clip lasting a few seconds. They struggle to keep a story consistent across many shots. Pandey told The Hindu BusinessLine that current tools often lose context and can't go much past eight to 10 seconds. He said Dolphin addresses this with multi-shot technology for longer videos and an agentic workflow that strings the work together.10 One write-up described the familiar failure in concrete terms: a character's coat changes colour by the third scene, a window appears in a kitchen, and the lead actor's face slowly turns into someone else's.2

Eloelo says its fix is a proprietary system for multi-shot planning and continuity. It is meant to keep characters, voices, wardrobes, props, locations and visual style stable across a whole production.16 Mint reports that creators build these elements as reusable references and carry them from shot to shot, so a project is handled as one piece of work, not a pile of separate clips.1 Above that sits what the company calls an agentic production layer. It works out what the creator wants, splits a longer video into shots and tasks, and coordinates generation while keeping context.36

The most ambitious part is Film Studio. It reads a script for its premise, tone, chronology and character arcs, then pulls out the characters, locations and props. From those it builds reusable visual, voice and wardrobe assets, which creators can review and lock before production starts.34 There are four ways to work in the product. Chat takes natural-language prompts. Templates offer guided formats. Playground gives shot-level control. Film Studio is the fourth. Editing and enhancement are built into all of them.24 Mint's account lists templates for ads, avatars, podcasts, lip-sync, tutorials and social content. It also describes an editor for cutting, merging and rearranging clips and adding voice, music and sound effects.1

"The model is becoming a component"

For anyone tracking multimodal AI, the most revealing line came from CTO Sagar Gaonkar. He told BusinessLine that "the model is becoming a component" and "the creative system is the product."10 In a longer version of the quote, he says creative software shouldn't be about picking a winning model. Instead it should be a layer that understands intent, breaks up the work, sends each part to the best model, and keeps continuity when the parts are put back together.5 Gadgets Now summarised the same idea: different models do better at different jobs, so Dolphin chooses among them task by task.4

The outlets disagree a little on what that means in practice. Several describe the continuity and planning layer as proprietary. Technosports says the studio pulls together multiple AI models for planning, voice synthesis, editing and Film Studio features.19 NewsBytes says the product "works with multiple AI models (even ChatGPT)."5 That probably mixes up two separate things. The more detailed reports say Dolphin supports the Model Context Protocol (MCP), which lets AI clients such as Claude, ChatGPT, Gemini and Cursor call Dolphin's image, video and voice tools from their own agent workflows.310 Put simply, MCP means outside assistants can use Dolphin as a tool. It does not mean Dolphin is built on ChatGPT. None of the coverage names the underlying video or image models Dolphin sends work to.

This matters because the strategy is sensible but leaves Dolphin exposed. Being model-agnostic protects Eloelo from the fast turnover in frontier video models. It also means the hard part of generation is handled by other companies, and any competitor can rent the same models. So the lasting value has to come from the planning, asset-locking and continuity layer. Whether that layer holds up in long, messy projects is exactly what the coverage hasn't tested yet.

Built from production volume, not a lab

The best argument for Dolphin is where it came from. Every major write-up repeats that Story TV has made more than 5,000 original dramas and that Eloelo's teams test over 100,000 creatives a month. The tool was first built to speed up that in-house work.36 Pandey said Dolphin grew out of fixing the company's own production problems at that scale, and that the company is now opening up the technology.6 Technosports also reports a claimed 15 billion minutes of total consumption across Eloelo's platforms.9

This is a meaningful change for the company. Eloelo started in July 2020 as a live social entertainment app. It mixed livestreams in regional languages with games like tambola and antakshari, and made money from virtual gifting, not ads.1219 In April 2025 it raised Rs 114.3 crore (about $13.5 million) in a Series B led by Play Ventures. It said part of that money would go toward deeper AI-powered creator tools.15 Entrackr estimates put its post-round valuation at about Rs 1,026 crore, or roughly $121 million. Filings showed losses more than doubled to Rs 99 crore in the fiscal year ending 2024.1719 At the time, Eloelo reported a $23 million revenue run rate and said it was aiming for $60 million by the end of 2025.15 The Dolphin coverage doesn't say whether it reached that target.

My reading: Dolphin is partly a product and partly a hedge. Micro-drama production is costly and needs huge volume. A tool that cuts Eloelo's own costs and can also be sold to others gives the company a second business that rides the same AI cost curve.

Where it fits in a crowded field

Dolphin isn't entering an empty market. Technosports lists Higgsfield, HeyGen, Kling, InVideo, Runway, Pika and Luma as competitors.89 Even in India, the field is busy. In June 2026 the government's IndiaAI Mission backed Avataar's launch of Varya. Avataar describes it as a distilled video model that cuts generation from 50 steps to 4. By its own internal benchmarks it costs ₹0.48 per second of video.29 Varya is a bet on making the model itself cheaper. Dolphin is a bet on the workflow built around models. Both approaches can succeed together, and Dolphin's model-agnostic design could in principle plug into cheaper Indian models like Varya.

The cheaper end of the market is crowded too. Tools for India pitch regional-language voiceovers, AI avatars and rupee pricing to creators and small businesses.2224 An AI grants guide sums up the 2026 situation well. The real question is no longer whether AI can make an impressive clip. It is whether a team can keep characters consistent, control the camera, get usable dialogue and secure clear rights across an entire project.26 Dolphin's pitch lines up almost exactly with that standard. The guide's last point, rights, is one that none of the Dolphin coverage touches.

The "content exporter" pitch

Pandey's pitch to the country is bold. He said AI will create "a world of infinite content." He wants India to move from being one of the biggest consumers of content to one of the biggest exporters, and to become a hub for the "Orange Economy."1 The company cites estimates that India's creative economy is about 2.5% of GDP and that about 80 million people work in its creator economy.410 It says lower production costs could let small teams and solo creators make content for audiences beyond their home markets.10

The argument holds together, and the ministry's presence at the launch suggests policymakers like it.6 But several things are still missing: no pricing, no named models, no independent comparisons with Kling or Runway, and no public examples of long-form output. Until those appear, Dolphin is a well-designed bet on the orchestration layer. It comes from a company with an unusual amount of real production experience. It has not yet been shown to solve AI video's continuity problem.