Video creation as code.

What Tellers includes for developers:

  • API for programmatic video and AI workflows.
  • In-house player for instant timeline rendering.
  • Open-source timeline format and CLI.
  • AI analysis, indexing, and search on your content.
  • Agent for video editing.
  • Unified abstraction over leading generative video models.
  • No video infrastructure to run.

Building video features from scratch is a long detour

Video is one of the most infrastructure-heavy features to build. Tellers exists so you don't have to.

Building video infrastructure takes months

FFmpeg pipelines, transcoding workers, storage, CDN, thumbnail generation, indexing, AI search. Building this from scratch before shipping any feature is a significant detour.

LLM pipelines need a reliable rendering layer

LLMs can generate video timelines, but they still need a deterministic renderer and a centralized layer to search, store, and manipulate media across mixed sources and codecs. Tellers ships open-source tools (CLI, timeline library) for LLMs to drive timelines, plus an in-house player for instant previews — so your LLM pipeline has something reliable to plug into.

GenAI models fragment your integration surface

Each new generative video model ships its own API, auth, output format, and custom billing and pricing model to reconcile. Maintaining six separate integrations is maintenance overhead that grows with the model landscape.

The developer surface area.

An API, a format, and a CLI. That's the entire integration surface — opinionated enough to be productive, open enough to fit any stack.

REST API

Send your media, prompts, scripts, or timelines. Tellers stores, indexes, searches, edits, and renders your videos. Asynchronous by design with webhook callbacks and streamed progress events.

Open-Source Timeline Format

tellers-timeline is an OTIO-based JSON format for describing video compositions, backed by an open-source Rust library. Predictable, version-controllable, and straightforward for LLMs to generate. The same format used internally by the Tellers app.

CLI

The Tellers CLI makes it trivial to upload, analyze, and index your local videos. An even simpler command triggers the video editing agent: tellers "Create a summary of my last holiday footage".

Infrastructure you don't have to build.

Beyond rendering, the Tellers platform handles video intelligence, playback, and model orchestration — all available through the same API.

Video Intelligence

Tellers automatically analyzes, stores, and indexes every video asset — extracting scenes, transcripts, objects, entities, and semantics. Visual search ("find the scene with the whiteboard"), transcript search ("find the part about pricing"), entity search ("find every clip of the CEO"), and semantic search all work out of the box. No separate search infrastructure to build.

Tellers Player

A proprietary player built for live timeline previews. Streams clips from multiple servers simultaneously, handles mixed codecs and resolutions in a single playback session, and renders HTML and image overlays directly on top of the timeline — including Picture-in-Picture. No pre-transcoding needed before preview.

GenAI Model Aggregator

The Tellers API is a zero-overhead abstraction layer over the leading generative video models. Switch models or blend outputs with a single config change — no vendor lock-in, no separate API keys, no per-model integration work.

How the API works

  1. Upload your media
    Push raw footage, audio, or existing assets to Tellers via the API or CLI. Every upload is automatically analyzed and indexed, so it's searchable and ready to edit.

  2. Request an edit
    Send a prompt, a structured timeline, or an LLM-generated tellers-timeline describing the video you want. One call kicks off the agent pipeline.

  3. Stream progress
    Consume server-sent events in real time as Tellers plans the edit, generates assets, and assembles the timeline — no polling required.

  4. Preview, tweak, and render
    Open an instant preview link in the Tellers Player, adjust the timeline if needed, then trigger the final render.

Illustrative — see API reference for the full schema.

All the major GenAI video models. One API.

Tellers acts as a zero-overhead abstraction layer. Specify the model per render, blend outputs, or let Tellers route based on your requirements. New models are added to the aggregator as they reach production quality — your integration code doesn't change.

What teams build with Tellers

  • SaaS video features: Add AI video generation to your product without building or maintaining video infrastructure. Your users get video; you make one API call.
  • LLM-driven content pipelines: Use an LLM to generate a tellers-timeline JSON from a blog post, transcript, or data feed. POST it to the API. Get a finished video. End-to-end automation.
  • Media company workflows: Ad hoc integration into production pipelines. Automate upload, synchronization, and analysis of media so every asset is preprocessed and indexed — ready the moment you ask the agent for an edit.

Frequently asked questions

What is the Tellers API?
The Tellers API is a REST API for programmatic video workflows. You upload media, send prompts, scripts, or timelines, and Tellers stores, indexes, searches, edits, and renders your videos. It handles AI asset generation, captioning, audio sync, transcoding, and delivery. Progress is streamed via server-sent events, and final deliveries are sent through webhook callbacks.

How do I add AI video generation to my SaaS product?
Integrate the Tellers REST API. Upload your users' media (or skip upload entirely and generate from stock footage via a prompt), then request an edit with a prompt or a tellers-timeline. Tellers streams progress, gives you an instant preview link, and renders the final video asynchronously. There is no video infrastructure to provision or manage on your side.

What is the tellers-timeline format?
tellers-timeline is an open-source, OTIO-based JSON format for describing video compositions, backed by an open-source Rust library. It defines scenes, assets, captions, timing, and output parameters in a structured, predictable schema. It is designed to be generated by LLMs, written by hand, or produced programmatically, and it is the same format used internally by the Tellers app.

What is the Tellers video intelligence layer?
Every video processed by Tellers is automatically analyzed, stored, and indexed — extracting scenes, transcripts, objects, entities, and semantic metadata. Out of the box you get visual search ("find the scene with the whiteboard"), transcript search ("find the part about pricing"), entity search ("find every clip of the CEO"), and semantic search, without building a separate search or tagging infrastructure.