# Media Studio: podcasts, dubbing and video from one brief Entry E12. Sector: Prospicience product for creators and media teams. Status: in production. Delivered: 2026-09. Source: https://ai.prospicience.in/work/media-studio Media Studio is a Prospicience platform that turns a single brief into a finished podcast episode, a dubbed version or a short video. It routes every step to the best of more than ten AI services by cost and quality, switches to a backup when one fails, and records what every file cost. ## The challenge Producing a podcast episode, a dubbed version or a short video meant stitching six different AI services together by hand: script, voices, music, dubbing and video each lived in a separate tool. When any one of them hit a limit, the whole job stopped. ## Why it mattered A stalled job lands on a client deadline, and every hand off between tools is time spent moving files instead of making content. Spend was scattered across separate vendor bills with no way to see what a single episode actually cost, which made pricing client work guesswork. ## What we built - Takes a brief and returns a finished episode: script, several AI voices with the accents you choose, background music and a final mix. - Routes each step to the best AI service for cost and quality, and switches to a backup automatically when one fails or hits a limit. - Dubs into other languages with three engines, generates video from text or images, and keeps a running tally of what each job cost. - Gives admins user management, usage figures, an activity log and key rotation from one dashboard. ## The result - Creators go from idea to a mixed, publishable episode in one place instead of six. - A rate limit at one vendor slows a job down rather than stopping it. - Every generated file carries its cost and the service that made it, so spend per job is visible at a glance. - Scripts, voices and dubbing cover Indian English, Hindi, Tamil, Telugu, Marathi, Bengali, Gujarati and more. ## Built for trust Every file is stamped with the AI service that made it and what it cost, so spend is traceable to the job. User permissions live apart from user profiles, and admins get an activity log and key rotation built in. ## AI at work Inside Media Studio, AI writes the script, voices each speaker in the accent you pick, adds the music, dubs into other languages and renders video. Creators choose the voices, preview the lines and decide what ships. What it produces: Podcast and story scripts in Indian languages, Multi-voice episodes mixed to MP3, Dubbed versions in other languages, Video from text or images, A cost and a source on every file. - Six separate tools become one job, so creators spend their time on the content instead of moving files between services. - When one AI service hits a limit, a backup picks up the step, so a deadline job slows down rather than stops. - Lines finished in the voice studio move straight into a full episode, so good work is reused rather than redone. - Every file is stored with what it cost and which service made it, so pricing the next job starts from real figures. How it works day to day: A creator enters a title, source material or brief. AI drafts the script, casts the voices, adds music and mixes the episode, and the creator previews and adjusts before publishing. Each step goes to the best available AI service and switches to a backup if one fails. ## Figures from the delivered system - AI services connected: 10+ - voice engines with backup: 5 - dubbing engines: 3 - languages: Indian English, Hindi, Tamil, Telugu, Marathi, Bengali, Gujarati, more ## The technology - Brief in, finished episode out: script, voices, music, final mix - Picks the best AI service for each step by cost and quality - Switches to a backup automatically when one fails or hits a limit - Dubbing into other languages, and video from text or images - A running tally of what each job cost AI models used: Gemini (Google), GPT-4o (OpenAI), ElevenLabs, Sarvam, Seedance, Veo, MiniMax, Replicate. ### Technical notes - More than ten AI providers behind one interface, routed per step rather than per project - Five voice engines in a failover chain with multi key rotation - Language aware script routing: Indic and code mixed to one model, English to another - Three dubbing engines, plus text to video and image to video paths - Provenance and cost recorded against every generated file - Permissions kept separate from user profiles, with an admin activity log and key rotation ## Questions ### What happens when an AI provider goes down mid job? The job carries on with a backup. Every step has a failover chain, so a rate limit at one vendor slows a job down instead of stopping it, even in the middle of a client deadline. ### Which Indian languages does the dubbing cover? Indian English, Hindi, Tamil, Telugu, Marathi, Bengali and Gujarati among others, across three dubbing engines. ### How do you control AI spend across ten providers? Every generated file records which service made it and what it cost, and admins see usage across the whole studio. Ten providers add up to one itemised view of spend, not ten separate mysteries. Capabilities: ai-media-generation, voice-and-multilingual-ai, ai-agents-and-automation.