Update title
EvidenceCast AI is now live: public app, API and demo video
Update post
EvidenceCast AI has reached an important milestone: the project is now publicly accessible for judging and review.
The latest version demonstrates the complete evidence-to-media workflow:
- research findings are converted into traceable evidence cards;
- a human reviewer approves, edits or rejects each claim;
- only approved evidence enters the storyboard and narration stages;
- Genblaze manages scene-level media orchestration and provenance;
- Backblaze B2 stores source records, reviews, prompts, provider results, failures, manifests and final delivery assets;
- subtitles, thumbnails, infographics and captioned video outputs are produced with explicit traceability to the approved source evidence.
The public application includes three synthetic fixtures covering community health, climate resilience and education. These fixtures contain no personal data and allow judges to explore the approved workflow without requiring paid provider credentials.
A major focus of the latest release is transparency. Provider failures are stored and reported rather than hidden. The reviewed local delivery remains clearly marked as:
provider_generated: false
This ensures that fallback media is never misrepresented as a successful provider-generated asset.
The project now includes:
- a public Streamlit application;
- a public FastAPI service;
- a public GitHub repository;
- Backblaze B2 workflow restoration;
- SHA-256 asset verification;
- scene-level regeneration with parent-child lineage;
- three approved demonstration fixtures; and
- a public demonstration video under three minutes.
Try EvidenceCast AI
Live application: https://evidencecast-web-buriro-2026.onrender.com/
Public API: https://evidencecast-api-buriro-2026.onrender.com/
Source code: https://github.com/buriro-ezekia/evidencecast-ai
Demonstration video: https://www.youtube.com/watch?v=tCU5KRakVxE
Log in or sign up for Devpost to join the conversation.