URL to video
URL to video is now available in Poko Motion. Paste a website URL and the agent captures the page, reads the real site content and visuals, then builds a motion video without needing a local repo.
Follow along as we build Poko Motion. Shipped updates, planned features, and product notes straight from the team.
URL to video is now available in Poko Motion. Paste a website URL and the agent captures the page, reads the real site content and visuals, then builds a motion video without needing a local repo.
Planned OpenAI-compatible BYOK support for Straico, OpenRouter, Cerebras, Gemini, and other hosted model providers so users can bring their own keys and choose the model stack that works best for them.
Planned BYOK support for voice providers like Cartesia, ElevenLabs, Sarvam, and similar APIs so users can connect their own speech generation accounts.
Planned upload flow for user media assets plus voice cloning support, making it easier to build videos around real footage, brand assets, and custom narration voices.
Planned sound design controls for background music, sound effects, ambience, and simple audio mixing so generated videos feel more finished.
Planned community space inside Poko where users can discuss workflows, share videos, ask questions, and learn from other creators.
Planned recording and editing workflow for influencers and YouTubers, with motion overlays, callouts, and AI-assisted edits layered on top of screen or video content.
Planned BYOK support for generation providers like Kie.ai, FAL.AI, and related media APIs for images, video, and other creative assets.
Planned cloud upload and sync so desktop projects can be backed up, restored, and continued across machines without manually moving workspace folders.
Planned version history for motion projects, so the first cut can be preserved and later edits can branch into new saved versions instead of overwriting the same result.
Planned web-based generation for website, PPT, and PDF to video flows. Desktop will remain the local repo-focused workflow, while non-repo inputs can run from the website.
Planned REST API and Model Context Protocol support so users and teams can connect external tools, knowledge sources, automations, and custom workflows directly into Poko Motion.
The desktop app now fetches the signed-in user profile from Clerk and shows account details plus sign out directly in the app sidebar.
Poko Motion now uses Opus 4.8 as the latest video-generation agent. The upgrade improves scene planning, visual density, code quality, and multi-step edits for polished motion videos.
Motion ads now default to a tighter 40-45 second format with higher animation density, stronger hooks, and faster pacing. Users can create a 45-second high-motion ad for approximately $3 in AI usage.
Added clearer product walkthrough and motion-ad guidance so users can understand what to upload, what the agent builds, how long videos should be, and how to iterate with chat edits.
Render your motion video directly to your Downloads folder. A live progress bar tracks frame capture and FFmpeg encoding in real time. Once done, reveal the file in Finder with one click.
Replaced the full-browser preview with a lightweight embedded HyperFrames player inside the app. Faster load, no more Chrome pop-ups, and seamless scrubbing without leaving the editor.
Group projects into named workspaces for agency-style multi-client management. Create, rename, delete workspaces and move projects between them. Gated behind an agency flag for now — billing-based gating coming soon.
Live display of accumulated agent cost, input/output tokens, cache hits, and model usage per project. Resets per project, updates in real time as the agent works.
Hand the agent a .pptx file. It extracts slides, preserves layouts and branding, adds cinematic motion and transitions, and renders a polished video. No manual slide recreation.
Select any PDF — pitch deck, whitepaper, product doc. The agent reads every page, writes narration, and produces animated slides with transitions. Page-by-page extraction with auto-generated scenes.
Switched the AI backbone from OpenAI to Claude (Anthropic). Significantly better code generation quality for video compositions, with thinking-budget controls and BYOK support for users who want to use their own API key.
Massively improved motion density: varied GSAP transitions, staggered reveals, parallax layers, camera-drift effects, and dynamic zoom pulses. Videos now look cinematic instead of slideshow-like.
Point at any project repo — a SaaS codebase, design system, or marketing repo. The agent reads your assets, extracts brand colors and fonts, writes a script, and generates a full motion video locally.
Natural language editing inside the studio. "Make scene 2 zoom slower." "Swap to a dark background." The AI agent edits video files live and the preview updates in real time.
Everything renders on your hardware — no cloud queue, no per-render fees. M-series Macs render a 30-second video in under 60 seconds. Bundled FFmpeg and headless browser, zero external dependencies after first run.
Refactored frontend into smaller focused components, consolidated icons, centralized formatters, extracted shared types. Fixed tool rows stuck in "running" state, markdown table rendering, and agent context retention across turns.
Usage-based billing for motion video generation. Track token consumption, manage credits, and view detailed cost breakdowns per project. Free tier included with every account.