AI-Assisted Video Captioning for Social Media
What it is for
- Adds captions to social media videos produced by the NYC Comptroller's office, making video content accessible to people who are deaf or hard of hearing.
- Supports information presentation by ensuring that social media videos from the Comptroller's office carry accurate caption text before publication.
- Run by
- Office of the Comptroller
- Where
- No fixed location
- Kept
- Not stated by the NYC Office of Technology and Innovation.
- Shared with
- Accountable organization, Vendor
What it is for
Mirage is a captioning tool used by the NYC Office of the Comptroller to automatically add captions to social media videos. Staff upload videos and the tool generates caption text that staff then review before publishing. The tool is used approximately once a week and human staff monitor its outputs before any video is made public.
What it collects and what happens to it
Data taken in
- Video taken for social media, which may contain audio of voices and visual depictions of people. The AI processes these signals to generate caption text but does not perform identity recognition.
Processing
- Converts spoken words in the social media video into caption text (speech-to-text transcription). The tool uses audio signals from the video to produce accurate captions.
What it does
- The system processes video input to extract and transcribe the spoken or textual content, converting audio/visual signals into structured caption text. Staff review and approve captions before publishing.
- The tool generates caption text to be overlaid on the video, creating new textual content that did not previously exist as a separate artifact. Human staff review all generated captions before publication.
Outputs
- Video output with captions added to the input video, ready for social media publishing after human staff review. The captions are new text generated by the AI tool.
Run by
- The NYC Office of the Comptroller's Communications Department deploys Mirage to caption social media videos produced by the Comptroller's office.
Built by
- Mirage is the off-the-shelf vendor whose captioning product is licensed and used by the Comptroller's Communications Department. Training data is proprietary to the vendor.
Kept for
Not stated by the NYC Office of Technology and Innovation.
Shared with
- The captioned video output is available to the NYC Office of the Comptroller's Communications Department, which reviews and publishes the final video on social media.
- Training data is proprietary to the vendor. The extent to which the vendor accesses processed video or caption data during tool operation is not specified in the register entry.
Stored
Not stated by the NYC Office of Technology and Innovation.
How to read the colours
Can it identify you?
- Anonymized data
- Data about people with the link to who is broken. Stripped of identifiers, blurred, aggregated, or noised so this system can’t reasonably tie a record back to an individual.
- Pseudonymous data
- Each person’s data is tied to a token (hash, ID, template) that lets this system recognise the same person across events, but the token itself doesn’t reveal a name. Reidentification is possible with extra information.
- Identifiable data
- The data either contains a direct identifier (name, address, account name, recognisable face or voice, plate number) or carries a token this system uses to look up legal identity during processing.
Who completes the loop?
- Human decides
- This mode suggests; a person decides what to do next. The AI is always advisory — a human is in the loop on every decision. Example: a triage tool ranks cases for a clinician who chooses which to see first.
- Human executes
- This mode decides; a person carries out the result. Example: an optimizer plans the day’s trash-collection routes, and drivers run them.
- Autonomous
- This mode decides and acts on its own. No person reviews each decision or carries out the resulting action.
Definitions from the DTPR standard. Amber is about your data, violet about who decides. The fuller the shape and the deeper the colour, the more identifying the data or the less a person is involved.
- AI registerNYC Automated Decision Systems — AI Use Cases Register (2025)NYC Office of the Comptroller, Communications Department. Tool name: Mirage. Year reported: 2025.
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- AI registerNYC AI Use Cases Register 2025
- Register entryPublished by the NYC Office of Technology and Innovation. Reference bb2aad80. This disclosure was drafted with AI assistance.Schema: ai@2026-05-06-beta
What you can do
Ask about this system
Questions go to the NYC Office of Technology and Innovation, not the vendor.
Your rights
- Right to Be Informed of AI UseMembers of the public should be aware that the NYC Office of the Comptroller uses an AI captioning tool (Mirage) to add captions to its social media videos. This disclosure is available through the NYC AI Use Cases Register.
- Right to Algorithmic TransparencyThe NYC Office of the Comptroller has disclosed that Mirage is an off-the-shelf captioning tool whose training data is proprietary to the vendor. Details about the underlying model and algorithm are not publicly available. Members of the public may contact the Comptroller's office for further information.
Risks and safeguards
- Societal & cultural harmInaccurate automated captions could spread misinformation or misrepresent official statements from the Comptroller's office.Safeguard: The tool is described as 'monitored' with human staff reviewing outputs before publication, approximately once a week, to catch and correct errors before any video is made public.