DNA (Dailies Note Assistant) is an open initiative under the Academy Software Foundation to create a platform to unify manual note taking, version context, and transcriptions. This empowers studios to build tools to enhance the accuracy and speed at which creative notes are passed to artsits for dailies sessions. It aims to explore how large-language-model-assisted transcription and summarization can meaningfully enhance the VFX dailies note-taking process, prioritizing practical utility and thoughtful support for production professionals. This is the first project proposed in the ASWF Machine Learning Workgroup
To get involved, see user stories, understand workflow and much more please visit the WIKI
The goals of this project are detailed in the original proposal
Today, review notes are usually taken by a single coordinator who has to follow the entirety of a meeting perfectly as it jumps around with different feedback associated with different media and tasks.
This can lead to:
- Lost context and feedback between reviews
- No standard for note taking
- Inconsistent version tracking
- Limited interoperability
- Poor AI usability
With DNA, review workflows become:
- Clear - notes tied to exact frames, versions, and context
- Continuous - feedback carries forward across iterations
- Connected - works across review tools, trackers, and pipelines
- AI-ready - structured data enables summaries, actions, and insights
DNA seeks to be an open-source, integratable tool that improves accuracy and efficiency in capturing review feedback without removing essential coordinator responsibilities.
A shared structure for:
- Who gave the note
- When it was given
- What frame/timecode it applies to
- Which version it belongs to
- Status (open, resolved, etc.)
- Relationships (replies, carry-forward)
Based on real production workflows:
- Live review Dailies sessions
- Multi-user collaboration
- Version iteration cycles
- Note carry-forward
Because notes are structured, you can:
- Turn transcripts into usable notes
- Auto-generate summaries
- Extract actionable tasks
- Track feedback across versions
Example Workflow:
1. Director gives notes during review
2. Notes are captured (manually or via transcription)
3. Notes are structured using DNA
4. Notes attach to a version + timecode
5. Teams track progress without re-copying anything
DNA is built for:
- Directors
- Supervisors
- Coordinators
- Pipeline engineers
- TDs
- Tool developers
- Studios designing heavy dailies review pipelines
- 🚧 Early-stage (actively shaping the standard)
- 🤝 Seeking industry feedback
- 🧪 Validating real-world workflows
- Give Feedback
- Share Workflows
- Highlight Painpoints
- Validate or challenge assumptions
- Ask a question:
- All can be done via ASWF Slack -- join for the #dna channel for the discussions about this project.
- Attend a meeting:
- Technical Steering Committee meetings are open to the public, bi-weekley on Mondays 12:30pm PST
- Calendar: Zoom Meeting
- Meeting Notes: HERE
- Report a bug:
- Submit an Issue: https://github.com/AcademySoftwareFoundation/dna/issues
- Contribute a Fix, Feature, or Improvement:
- Read the Contribution Guidelines and Code of Conduct
- Submit a Pull Request: https://github.com/AcademySoftwareFoundation/dna/pulls
- GitHub project page
- The DNA project was established by this proposal
- ASWF's Machine Learning Working Group proposal -- describes the purpose and scope of MLWG. Join the ASWF Slack -- join for the #wg-ml channel for the discussions about Machine Learning
Documentation, including quick start, architecture, and API docs, are available on: QUICKSTART.md
For information on how to contribute to DNA please visit: CONTRIBUTING.md
The DNA project is part of the Academy Software Foundation, a part of the Linux Foundation formed in collaboration with the Academy of Motion Picture Arts and Sciences. The Technical Charter and Project Governance explain how the project is run, who makes decisions, etc. Please be aware of our Code of Conduct.
This project is (c) Copyright Contributors to the Dailies Notes Assistant project.
For original code, we use the Apache-2.0 license, and for documentation, the Creative Commons Attribution 4.0 Unported License.