Eloelo opens Dolphin AI, an agentic studio for multi-shot video
The Bengaluru company's platform plans a script into scenes, keeps characters and wardrobes steady between shots, and plugs into Claude and ChatGPT over MCP.
Key takeaways
- Eloelo Group unveiled Dolphin AI, an agentic video production platform, at FICCI FRAMES 2026, and it is available through an early-access programme with invite codes.
- Moneycontrol reports the platform routes work across more than 25 image and video generation models, with a Playground for manual model and reference selection.
- Dolphin AI's Film Studio analyses a script for characters, locations, props, tone and chronology and builds reusable visual, voice and wardrobe references to hold continuity across scenes and shots.
- Eloelo says its Story TV arm has produced more than 5,000 original dramas and tests over 100,000 creatives a month, and Dolphin AI came out of that production work.
Eloelo Group unveiled Dolphin AI, an agentic video production platform that takes a brief or a script through planning, generation and editing inside one workspace, at FICCI FRAMES 2026. It is available now through an early-access programme.
The IndianWeb2 report carries Eloelo's own framing of where the tool came from: "Video is at the heart of what we do at Eloelo Group. We have produced more than 5,000 dramas and test over 100,000 creatives every month," said Saurabh Pandey, founder and chief executive of Eloelo Group. "Dolphin came from solving our own production problems at that scale."
The platform's central part is a routing layer. Moneycontrol reports that Dolphin AI works across more than 25 image and video generation models and breaks a brief into individual production tasks, sending each one to a model suited to it. A Playground lets a user pick models and references by hand, and Chat and Templates cover guided workflows.
Film Studio, which the company's own site lists as coming soon, is aimed at the problem that stops most multi-scene work: continuity between shots. Moneycontrol reports that it analyses a script for characters, locations, props, tone and chronology, builds reusable visual, voice and wardrobe references, then generates scenes and shots from them, keeping characters, voices, wardrobes, props and locations consistent as a project runs. Editing, voice, music, sound effects and upscaling sit in the same workspace.
Dolphin also plugs into other agents. Its own MCP page publishes a server URL for Claude, ChatGPT, Gemini, Cursor and Perplexity, and lists image models including Nano Banana Pro and GPT Image 2, and video models including Seedance 2.5, Kling v3 and Gemini Omni 1.1 Flash. The company says a long-form mode turns one script and one reference image into a talking video of 30 seconds to several minutes, always in 9:16.
Access is by invite. Creators can join the waitlist at trydolphin.ai or access the platform with an invite code, the company's release says, and Moneycontrol reports early access is running through invite codes.
Early access is by request at trydolphin.ai; no pricing or general release date has been published.
Sources
- moneycontrol.com - source
- indianweb2.com - source
- gadgetsnow.indiatimes.com - source
- trydolphin.ai - source
- trydolphin.ai - source