Audio Reports

Audio Reports turn a generated intelligence report into an MP3 narration. The feature is available in active Pro and Elite workspaces and is currently labeled Alpha in the profile editor.

Enable narration for a profile

The voice library groups available voices into News & Broadcast, Professional, Narrative & Documentary, and Casual & Conversational. The stored catalog can include English and Spanish voices with masculine, feminine, or neutral labels. Availability is data-driven, so the exact voice list can change.

What happens during generation

Y2 prepares a speech-specific version of the report before calling the text-to-speech provider:

• It removes the trailing references section, citation markers, HTML, and Markdown formatting.
• It normalizes headings, punctuation, common entities, numbers, currency, acronyms, and tickers for spoken output.
• It adds the assigned branding template's audio intro, before-report audio block, after-report audio block, and outro. When no value is configured, Y2 branding defaults apply.
• It applies the assigned pronunciation dictionary only when that dictionary has a provider ID and a synced status.
• It generates an MP3 with Cartesia Sonic-3 at 44.1 kHz and 128 kbps, then stores the file and its metadata in Convex.

Audio output is limited to 20 MB. Generation is also skipped when its estimated cost would push the report above a configured per-report budget.

The stored duration is an estimate based on the prepared text, not a measurement read from the finished MP3. The in-app player replaces it with the browser-reported media duration after the file loads.

Create a pronunciation dictionary

Pronunciation dictionaries are also an Alpha feature. Pro workspaces can store up to 10 and Elite workspaces up to 100. Each dictionary can contain up to 200 entries.

Editing entries moves the dictionary back to pending before it is synced again. If a selected dictionary is pending, errored, deleted, or outside the profile's tenant scope, Y2 does not apply it during generation.

Customize spoken branding

The InfoOps → Audio page also exposes the intro and outro for each branding template in the active workspace. Use {name} to insert the profile name.

The before- and after-report audio fields live in the full Branding Templates editor. A template must be explicitly assigned to the profile for its custom values to apply.

Play and deliver audio

When audio is available, the report page displays Listen. The bottom-sheet player supports play and pause, seeking, download, and playback rates of 0.75x, 1x, 1.25x, 1.5x, and 2x. Voice and duration metadata are displayed when available.

Delivery surfaces expose audio differently:

Audio files are removed when their owning report, profile, workspace, or account is deleted and can also be affected by storage cleanup. Do not treat a returned storage URL as a permanent archive.

Retrieve audio through the API

Use an API key with the reports:audio scope, or complete the endpoint's x402 flow. The path uses the report's public API ID.

curl --request GET \
  --url "https://api.y2.dev/api/v1/reports/REPORT_ID/audio" \
  --header "Authorization: Bearer YOUR_API_KEY"

{
  "data": {
    "url": "https://example.convex.cloud/api/storage/…",
    "duration": 847,
    "durationSeconds": 847,
    "durationFormatted": "00:14:07",
    "format": "mp3",
    "mimeType": "audio/mpeg",
    "fileSizeBytes": 13420800
  }
}

duration is retained for compatibility and is deprecated; use durationSeconds. Add ?redirect=true, or request an audio media type with the Accept header, for a 302 response to the stored file URL.

The related GET /api/v1/reports/{reportId}/audio-text endpoint requires reports:read and returns the speech-preprocessed text, character count, and estimated duration without requiring that an MP3 already exists. See the generated Reports reference for canonical schemas, errors, rate limits, and x402 pricing.

Troubleshoot audio

Next steps