OpenAI TTS requires AI-voice disclosure: burn it with Sume cues

OpenAI's text-to-speech guide states that you must clearly disclose to end users that the voice they are hearing is AI-generated. If your video uses a synthetic voice from any provider, a visible notice is a simple way to do it. Sume's POST /v1/video-captions accepts cues, which burns your own text onto the video with start and end times and no speech recognition.

What the OpenAI guide and Sume docs say

This is not legal advice; it is what the vendor pages state.

Burn a notice

Send one cue that spans the opening seconds. Only one of script_text, words, cues or segments is allowed in a request.

import os, requests

r = requests.post(
    "https://api.sume.com/v1/video-captions",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
             "Idempotency-Key": "ai-voice-notice-001"},
    json={"video_url": "https://example.com/narrated.mp4",
          "style": "black-outline",
          "cues": [{"text": "Narration: AI-generated voice",
                    "start": 0.0, "end": 4.0}]},
)
print(r.status_code, r.json())

Placement and wording

Pick a style that stays readable on your footage, and keep the notice on screen long enough to read. A notice in the post description or on the page that hosts the video is an additional option; follow the rules of each platform you publish on.

Worked example

Think of the notice as part of your template.

• Add a single cue near the start of each narrated video.
• Use the same wording and style every time.
• Keep the cue text in your repo so changes are reviewed.

Checklist before you commit

A burned-in notice cannot be removed by a viewer; if you need an optional notice, use platform fields as well. This post does not give legal advice.

• Ask counsel what your market requires.
• Check each platform's AI-label settings.
• Keep a record of which videos used synthetic voices.