Export Format Specification v0.9.

This document describes the structure of the export bundle a TellSophie account holder receives when they request an export. Today that request is fulfilled by hand within five business days. Version 0.9 is provisional: the schema is locked.

We publish the spec so the export commitment is inspectable. Any future tool, service, or self-hosted player can read TellSophie collections without our cooperation by following this spec.

Related: Trust commitments.

Bundle structure.

The export is a single .zip file. Top-level layout:

• manifest.json — required. Bundle metadata, schema version, child id, timezone, and a content listing.
• recordings/ — directory of media files, one per recording. Filenames match the id field in the per-recording sidecar.
• recordings/<id>.json — sidecar JSON for each media file. Stored alongside the media (not in a separate sidecars/ directory) so a tool can match a sidecar to its media by base filename.
• transcripts/ — directory of plaintext transcripts, one per recording with a transcript. Same id-based filename convention.
• README.txt — human-readable note explaining what the bundle is and how to read it.

manifest.json.

Top-level fields, all required unless noted:

• schema_version: string. The export-format spec version this bundle conforms to. v0.9 emits "0.9".
• bundle_id: string. ULID. Unique per export.
• exported_at: string. ISO 8601 UTC timestamp.
• child_display_name: string. The display name as configured by the account holder at export time.
• child_id: string. ULID. Stable across exports.
• timezone: string. IANA tz database name (e.g. America/Los_Angeles).
• recording_count: integer.
• recordings: array of objects with id, created_at, duration_ms, format, contributor_display_name, and prompt_id fields. The full per-recording detail lives in the sidecar JSON; this array is a fast-index for tools that want to enumerate without opening each sidecar.

Per-recording sidecar.

recordings/<id>.json carries:

• id: string. Matches the media filename.
• created_at: string. ISO 8601 UTC.
• duration_ms: integer.
• format: string. One of audio/m4a, video/mp4. Audio uses AAC at 128kbps; video uses H.264 at the device’s native resolution capped at 1080p.
• contributor_display_name: string.
• contributor_id: string. ULID. Stable across the contributor’s recordings.
• prompt_id: string. The prompt the recording responded to. Stable IDs published in a separate prompt-catalog version that increments alongside the schema.
• prompt_text: string. The prompt text rendered to the contributor at recording time. Captured here for archival even though prompt_id is the stable reference.
• transcript_path: string or null. Path within the bundle relative to the bundle root.

Standards used.

• Container format: ZIP per APPNOTE.txt 6.3.x.
• Audio encoding: AAC-LC, 128 kbps, 44.1 kHz, in MPEG-4 container (.m4a).
• Video encoding: H.264 baseline profile up to 1080p, AAC audio track, in MPEG-4 container (.mp4).
• Transcripts: UTF-8 plaintext, no timestamps in v0.9 (timestamped transcripts target v1.0 alongside the iOS app launch).
• IDs: ULID per the ULID spec.
• Timestamps: ISO 8601 UTC with millisecond precision.

What changes from v0.9 to v1.0.

Version 1.0 will lock these additions:

• transcript_segments array on the sidecar with per-utterance timestamps.
• A signature.txt file at the bundle root carrying a detached cryptographic signature over manifest.json so a future tool can verify the bundle was emitted by TellSophie infrastructure (independent of the bundle’s file integrity).
• A prompts.json catalog snapshot embedded in the bundle so a future tool does not need network access to resolve prompt_id references.

The schema fields above will not change between v0.9 and v1.0. Additions only.

Questions.

If you are building a tool against this spec and find an ambiguity, email privacy@tellsophie.com. We will publish clarifications on this page.