Colossyan?

NOT REALLYA prototype of the interface is possible, but the durable product depends on difficult systems beyond the UI. Build a safer adjacent workflow instead.

BUILD FIRST

A non-sensitive or read-only version of the core experience
A manual review step wherever correctness or trust matters
A narrow companion tool that removes a real piece of work

LEAVE FOR LATER

Regulated decisions, payments, or irreversible actions
Claims of correctness, compliance, or uptime that the MVP cannot support
Sensitive integrations until security, support, and auditability are ready

RESEARCH NOTES

What this tool actually does.

Checked 2026-09-06

Colossyan is an AI video generation platform for corporate training and internal communications. Its flagship workflow starts with a source document (PDF, PowerPoint, Word, Google Doc, Notion, URL, or pre-written script) which the "Cora" AI agent converts into an editable, cited video scene plan for user approval before rendering. The platform provides 300+ photorealistic AI avatars speaking 100+ languages with lip-sync and gestures, plus custom avatar and voice cloning from short recordings. Editing covers script rewriting, AI voiceovers, auto-subtitles, screen recording, and templates. A companion LMS component ("Colossyan Learn") supports course outlines, in-video quizzes, branching scenarios, completion analytics, and SCORM 1.2/2004 export. Multi-avatar conversations enable dialogue and roleplay-based training. The product is primarily a video/course authoring tool rather than a pure

SOURCES

RESEARCHED BUILD PROMPT

Give this to your coding agent.

Build a focused open-source replacement (no Colossyan branding, logo, marketing copy, or UI layout) for the core Colossyan training-video workflow: document-to-video via AI agent. Workflow: user uploads a PDF or pastes a URL/script; a backend agent (LLM pipeline: LangChain or direct OpenAI/Anthropic calls) extracts key points, breaks them into 5–15 scenes, each with script text, source citation, narrator voice preset, and previews an editable plan for user approval before any render. UI: Next.js two-pane editor — left scene list with citations, right 16:9 preview pane. Nothing renders until the user approves. Rendering approach: per scene, TTS (Piper local fallback or ElevenLabs cloud adapter) generates WAV; Rhubarb Lip Sync derives mouth shapes from the audio; a Canvas/WebRTC overlay animates a static avatar portrait; ffmpeg concats scenes into MP4 with optional burned-in SRT subtitles. Avatars: 4–6 built-in openly licensed portraits plus custom image upload; no real-time generative video. Out of scope: multi-avatar conversations, branching quizzes, LMS analytics dashboards, video-to-video translation, collaboration/version control, voice cloning, streaming generation, and commercial avatar libraries. Deliver a monorepo with Next.js frontend, FastAPI backend, SQLite for scene drafts, ffmpeg CLI integration, and a README covering setup, required API keys, known limitations (avatar realism far below commercial platforms, single-avatar scenes only, no cloud GPU rendering), and an ethics note on requiring consent for custom portrait/voice uploads.

Do not copy the original brand, proprietary assets, or UI; create an original visual system.

Visit the official product page