Image and video decision
Hong Xuan Kow
A skilled developer can build a narrow AI-driven video generator (prompt→render) in ~30 hours using third-party APIs and FFmpeg, but reproducing Koe's full product (managed premium models, credits marketplace, priority render infra, and enterprise features) is substantially more work.
Visit website↗Open-source builds that already do this
Every project below is open source and already does this job today. Fork one, self-host it, or take the parts you need — the build prompt further down assumes an empty file, and this is the shortcut past that. Licences differ; check the one on each card before you ship.
What a replacement has to do
- Provide a script or prompt → generate storyboard/shot list → synthesize clips (video frames, TTS/voice) via external generative APIs → assemble and render timeline with FFmpeg → produce captions and upscaled export.
What it still won’t have
- Proprietary premium models and premium-model access
- Built-in credit packs and integrated credit marketplace
- Priority render queue and guaranteed fast rendering
- Custom LoRA character training and enterprise features (SSO, SLA, dedicated support)
- Seamless 4K pro-quality rendering with managed infra and scaling
What remains hard
- Product polish and ongoing maintenance
First-year cost
Keep paying
Paying is—cheaper in year one.
On cash alone, building overtakes the subscription at 9 seats.
Money you would actually spend
Time you would spend
—
What you would spend
What we assumed
The verdict above measures whether you could build it. This one is only about money.
Runnable build prompt
Build a minimal AI video generation web app using Node.js + Express, React frontend, Postgres for projects, and FFmpeg for assembly. In scope: accept a text script or prompt, parse it into shots, call external generative APIs for frames/images and TTS, assemble timeline with FFmpeg, generate sidecar subtitles, support 720p–1080p export, user signup, and per-user credit tracking. Out of scope: custom model training, multi-tenant priority queues, enterprise SSO, and a marketplace. Include error handling for API failures, retry/backoff, input validation, and unit/integration tests for API handlers and render pipeline.
How we checked
How the score was reached
- Partly verdict base52
- An open-source build was found+5
- 4 cited sources+3
- Price verified on pricing page+3
- Evidence score63
The base comes from the verdict. Everything under it is a check that either happened or did not, and each one is a fact frozen in this record rather than a judgement made at render time — so the same evidence always produces the same number.
How scoring works →Cited sources · 4
Every page the run actually retrieved.
- official productKoe — Make any video you can imagine.
- official pricingPricing · Koe
- open sourcemodelscope/FunClip
- open sourceOpenShot/openshot-qt
Integrity checks
What held up, and what did not.
