Inspiration
What it does
Inspiration
Creators often make one strong long-form video, then lose hours resizing it, rewriting captions, preparing covers, and manually organizing files for each platform. We wanted to turn that repetitive handoff work into one clear, creator-controlled workflow.
What it does
OneSource turns a creator-owned 16:9 source video into a ready-to-review multi-platform delivery pack:
- 9:16 vertical video for Douyin and WeChat Channels
- 1:1 square video for Xiaohongshu
- 16:9 landscape video for Bilibili and YouTube
- Platform-specific titles, descriptions, and hashtags
- A cover asset, publishing-plan CSV, and machine-readable manifest
- Optional reviewed multi-platform publishing commands
Instead of automatically publishing without oversight, OneSource keeps creators in control: they review the crop, copy, cover, schedule, and destination account before any publishing action.
How we built it
We built OneSource as a local-first Python web app.
FFmpeg renders the three format variants locally. The app extracts a key frame and, when an OpenAI API key is configured, uses OpenAI to estimate the subject's horizontal position, generate platform-specific copy, and create a cover image. Without an API key, the full pipeline still works with center crops, editable local copy templates, and an SVG cover fallback.
The final output is packaged into a ZIP file containing videos, copy, cover assets, a publishing plan, and a manifest. We also export an editable Premiere-compatible XML workflow for creators who want to refine the final edit.
Challenges we faced
The biggest challenge was balancing automation with creative control. A fully automatic crop can fail when a subject moves, so OneSource treats AI focus detection as a starting point—not a final editorial decision.
Another challenge was platform fragmentation. Each platform needs different aspect ratios, copy styles, metadata, and publishing flows. We solved this by separating asset generation from publishing: creators first receive a transparent, local delivery pack, then optionally review platform commands before sending.
We also designed the project to remain useful without external services. FFmpeg handles real local video conversion, while OpenAI features are optional enhancements rather than a requirement for the core workflow.
What we learned
We learned that the most valuable creator automation is not “one-click magic.” It is a workflow that removes repetitive production work while making review points explicit.
OneSource demonstrates that a single source video can become a structured, reusable content package—ready for different audiences, formats, and publishing schedules.
Log in or sign up for Devpost to join the conversation.