Tessact: product evidence and build options

The verdict

Run the pilot. Buy the analysis, never the search or the database. Treat Tessact as a metadata supplier that writes JSON into our CMS, and hold the line on three things: cross-account S3 role, a real export API, and a no-training clause.

The prize the tester spotted is real and bigger than reported — the AI output is a rich, timecoded structural model of each sampled asset, not just tags. But every product feature that matters to a subscriber (semantic search, mood/theme browse, skip-credits, better "more like this") is unlocked by holding that data in our CMS + Typesense, which we already run. If instead we call Tessact's search from the apps or leave the metadata living in their workspace, we rent our own discovery layer forever. The corrected extraction client retrieves every page of 13 discovered resource paths across eight sample assets; one face-lane request returns 409 for audio. Some resources report unavailable, and complete export rights, service authentication and long-term access require a contract. Their S3 connection currently asks for access keys.

127

successful GET pages; 103 of 104 resource slots returned 200

237 ms warm

search latency (not 15 s) · 2.7–4.7 s cold

Bengali

captions are script, not romanized

bulk-export or CMS-write paths observed

01What Tessact actually produces

The tester saw tags and a synopsis. The live platform produces a timecoded structural model for the eight analysed samples. Thirteen paths were discovered. Their responses are paginated and can report partial or unavailable states; the audio face lane returned 409. This is what raises both the value and the lock-in stakes.

Resource	What it holds	Per 25-min ep	Unlocks for us
`summary`	genres, moods, themes, languages, EN synopsis (short + long), editorial title, ranked top moments	1	mood / theme / setting fields, ai_synopsis
`scenes`	per-scene mood, setting, synopsis, character names, timecodes	15	scene_index, semantic search fuel
`segments`	titled chapters: purpose + synopsis + timecodes	5	chapters / jump-to-section
`moments`	ranked highlights, scored on action/emotion/conflict/decision	4	key_moments, hover previews, promo shortlist
`signals`	partial-state editorial exclusions: credits / intro / recap ranges with timecodes	19	Skip Credits + Skip Intro
`dialogue`	Bengali-script cues + romanized + speaker label + timecodes	217	captions, search index
`settings`	place/object facets with density + first/last-seen	72	setting_place / setting_period
`people`	separate face job: groups, sparse identity suggestions, confidence and source revision	106 groups; 1 assigned	candidate cast evidence, with human identity review
`timeline`	composition segments, story beats, moment markers	1	player chapters & markers

The sampled output is promising. For Indubala Bhaater Hotel E01 it wrote a plausible synopsis, tagged it nostalgic / poignant / melancholic and memory and nostalgia / food and culinary heritage / widowhood and survival, timecoded "Laxmi's Tragic Fate on the Railway Tracks", and offered a face-identity suggestion. Across Indubala’s 106 reviewable face groups, one was assigned an identity, six had suggestions and 99 remained unknown. The Bengali dialogue came out in script (সোনা মুগের ডাল, বেগুন ভাজা), not romanized.

Coverage across all 8 sample titles (measured)

Title	Runtime	Scenes	Segments	Moments	Dialogue	Settings
Byomkesh E01	58m	25	21	12	421	126
Indubala Bhaater Hotel E01	25m	15	5	4	217	72
Jamai Bodol	14m	12	7	4	139	51
Meye Bela E01–E04	13–16m	6–13	4–9	3–6	92–156	40–56
Dotta E01 (FM audio)	29m	5	4	5	149	4

Pulled read-only for all 8 assets, ~4 MB JSON total. Audio (Dotta) gets dialogue + moods but no visual settings, as expected.

02The tester's read, checked against the live platform

The tester's instincts were right on the big picture and wrong on several specifics. The specifics matter, because two of them (captions, search speed) would have argued against a deal that's actually worth doing.

Claim in the plan	Verified	What we actually saw
Captions are romanized Bengali	wrong	Captions are Bengali script (মনিরুল।). A romanized field exists alongside, not instead.
Search took ~15 s	wrong at rest	Warm 237–960 ms, cold 2.7–4.7 s. Bengali-script queries work. 15 s was a cold first hit.
Closing credits at 3:23	mislabelled	3:23 is the opening credits. It segments both; real skip target is per-episode (Indubala 22:59).
Wants S3 keys, not a role	confirmed	Connection form has access key + secret key only. No role-ARN / ExternalId field anywhere.
No CMS connector	confirmed	Nothing writes to a CMS. It only ever proposes; we pull.
Only English tags	confirmed	Genres/moods/themes/synopses are English. Dialogue is Bengali script.
Metadata Templates empty	confirmed	Templates, content profiles, connections, automations all empty — greenfield.
Push results to an address we nominate	not demonstrated	The observed webhook form limits destinations to Slack and Zapier. The documented payload example links back to Tessact; an inline results payload and delivery guarantee have not been shown. See §04.

03Own vs rent — where the lock-in lives

The same data can make us independent or dependent, depending on where it lives and who runs the query. This is the decision with the longest tail.

What we own (the plan)

Tessact analyses → results land in our CMS as pending fields → content team approves → approved data feeds our Typesense and the apps.

CMS stays source of truth.
Search runs on infra we already operate.
Swapping Tessact out later is a supplier change, not a rebuild.
We can enrich, re-rank, and combine with watch data freely.

What locks us in (avoid)

Apps call tessact.com search directly · metadata lives only in their workspace · we depend on their webhook to ever learn a job finished.

Pay per query, indefinitely, at their latency.
Discovery — a core product function — is a rented service.
Their derived assets sit in their AWS account.
No contractual export = data hostage on renewal.

our S3 (masters) → Tessact (analyses) → OUR CMS (pending review) │ approved ├→ OUR Typesense (search index) └→ apps (browse rows, skip buttons) Tessact only ever proposes. It never writes to the CMS, and the apps never call it.

A narrow extraction path is proven. No bulk-export route was found. The first client saved only initial pages. The corrected client saved 127 pages across eight assets: 103 resource slots returned 200 and the audio face lane returned 409. Some 200 responses say partial or unavailable; embeddings, all derived media, revision history and every possible resource were not proven exportable. The first capture read library page 1; that workspace reported one page and eight total assets. The export client now follows library pages, expiring Content Intelligence cursors and face offsets. Contractually we are not covered yet: their public site terms are a 2022 visitor document with no data clause, and the retrieved AWS Marketplace contract has derived-data terms. Direct Enterprise contract terms remain unseen. That gap is the single most important thing to close before committing.

04Integration shape — we pull, they don't push

The plan assumed Tessact pushes finished metadata to an endpoint we choose. It doesn't, as built. This changes the wiring.

Webhook: An on_file_analyzing_completed trigger exists. The observed form host-allowlists hooks.slack.com and hooks.zapier.com; public docs show a URL to metadata in the sample payload. Backend enforcement, inline results, signing and retry are unverified.
Reality: We poll the REST API on our own schedule (exactly what the export client does). Polling is feasible for this small workspace. Service credentials, pagination, rate limits, backfill and endpoint stability still need vendor agreement.
Auth: Login sets a 7-day session cookie; the API takes a 15-minute bearer refreshed from the session. Standard.
Hosts: The paying Enterprise workspace is served from dev-api.tessact.com and a Cloud Run …-ai-dev service. Prod on dev-named hosts — a maturity signal, noted in §06.

05Fit with our stack

We already have the pieces to own discovery. Tessact fills the one gap we can't fill ourselves — reading the video — and nothing more.

Search: Typesense is live on EKS (3 replicas) with a semantic path (e5-small, 384-dim) already coded behind a flag. Feeding it Tessact's synopsis + scenes + dialogue is the intended use. Keep search ours. Qdrant / ClickHouse-vector are worth it only if Typesense's built-in embeddings prove too tight in its current 1 Gi pods — decide on measurement, don't pre-buy.
Skip buttons: The web + KMP players already read skip-intro / play-next fields. Tessact's signals can feed existing intro and play-next timing after editorial review. A distinct Skip Credits control needs client work; KMP marker mapping, cast and offline paths also need fixes.
Content model: Mongo already has mood[] (with a bn/hi/en lookup) and per-language synopsis fields; mood just isn't indexed yet. Theme, scenes, key-moments, structured warnings are new fields. Approval workflow exists at title level — we'd add a per-field "AI-suggested / pending" layer.
Recommendations: "More like this" is manual curation today. Approved themes and moods can supply candidate titles, but ranking and editorial controls still need design and measurement.

06Gaps & red flags

Security

Long-lived S3 access keys, not a cross-account role — the connection UI has no role option at all.
No SOC 2 / ISO 27001 report or sub-processor list was found in the public material; request the current documents and identify which AI providers process content.
API 404s return a ~372 KB framework error page — verbose error surface on the prod API host.
Derived crops sit in a Tessact-owned AWS account.

Contract & data

Public terms (2022) allow account deletion "without warning", exclude liability for data loss, say nothing on AI training or retention.
The retrieved AWS Marketplace contract contains derived-data return terms; direct Enterprise MSA/DPA terms remain unseen.
Masters >10 Mb/s are transcoded; where those copies live isn't disclosed. Our 5–13 GB files all qualify.

Vendor maturity

Public databases report ~23 staff and a 2018 seed round. The Enterprise workspace uses a dev-api hostname; the name alone does not establish its deployment controls. No public changelog entry was found after Nov 2025.
Named clients come only from Tessact's own press release.
Single sign-on appears only in unpublished docs; the observed login presents email / OTP / Google.

Bengali quality (open)

Captions in script — good. But diarization is generic "Speaker 1/2", no identity.
Tags/synopsis English-only; we'd translate the fixed vocabulary (~100 terms) ourselves.
Measured accuracy on our catalogue still unverified beyond the sample.

07What to demand before committing

The already-loaded sample supports an editorial quality review. A wider pilot needs a safe media-access plan, data terms and a measured cost/throughput proposal. These are the terms that decide whether we ever depend on them.

Read-only S3 via cross-account IAM role + ExternalId, scoped to a proxy prefix — not access keys, not our masters.
A documented, versioned JSON results API + bulk export — contractual, not the current dev-api.
Signed webhooks to our own endpoint, dropping the Slack/Zapier allowlist — or accept that we poll.
A written no-training clause and a sub-processor list (which providers see our content, in which region).
Measured Bengali accuracy on our content — script, romanization, diarization — over the 12–15 title pilot.
Where transcoded copies of our masters are stored, and deletion on termination.
What Content Profiles do in practice — the UI names cast, roles, structure and reference images, but the list is empty and the sampled profile context says not_generated. Demonstrate whether a profile improves consistency across episodes.
Work to our fixed vocabulary (moods/themes/settings), and output in Bengali as well as English.

08Recommendation

Yes to the pilot. The metadata is worth having and Bengali-script dialogue is present in the sampled output. Its accuracy needs a labelled evaluation. Buy it as an analysis supplier that drops JSON into our CMS; keep the search, the index, and the source of truth on our side. Make the cross-account role, the export API, and the no-training clause conditions of any paid contract — not nice-to-haves.

Effort: reviewing the eight existing outputs can be low engineering. A wider pilot needs safe media access and a measured throughput/cost proposal. Production use needs CMS suggestion fields and review, a supported ingest API, Typesense indexing and player work across platforms.

09Platform offer, observation and inference

These are different kinds of evidence. An enabled route or a marketing promise is not proof that a workflow is ready for this workspace.

Area	Platform offers or claims	Observed here	Inference or limit
Library and editing	Media asset management, metadata, review, editor, remixes, publishing and automation.	Library assets, technical fields, scrubs, review/AI routes, workspace roles and flags.	Editor, remix and publishing quality was not benchmarked; publishing and live-stream flags were off.
Content Intelligence	Scene understanding and searchable narrative structure.	Eight assets exposed 13 named resource paths; one audio face request returned 409. Other paths yielded paginated JSON: summary, scenes, segments, moments, quotes, dialogue, settings, people, timeline, translations, signals and editable summary/subtitle documents.	A 200 response may say unavailable. Translation, compliance, canonical people and speaker identity were unavailable on all eight.
Custom metadata	Templates, 13 field types, AI fill and spreadsheet export.	Template list empty. “New template” opens a name-only dialog; no template was created. Autofill API surfaced but was not invoked.	Closed vocabulary enforcement and Bengali output remain unproven.
Profiles and people	Content Profiles and face/person discovery.	The separate face lane yielded 473 reviewable groups across seven videos, with 16 assigned identities and 29 suggestions. Profile list empty; the page says profiles describe cast, roles, structure and reference images. Sampled profile context says `not_generated`.	Face suggestions are not canonical cast records. Profile effect is unknown.
Connections	S3 sync and completion automations.	Key-based S3 form, empty connection/automation lists, trigger/action catalogue.	Read-only cross-account role, direct Hoichoi webhook delivery and stable API authentication need vendor proof.

The internal evidence ledger, field-level coverage, source paths and unresolved questions are in the private Tessact assessment under ~/Tools/ExternalRetreival/tessact/ASSESSMENT-2026-09-28.md. The original DOCX's product proposal is treated as a hypothesis; the eight-asset export is a sample, not an accuracy benchmark.

10If we built the useful part ourselves

The target is Hoichoi metadata and discovery, rather than a replacement for Tessact's editor, media library and social publishing suite.

We do not need to train a video foundation model to begin. Tessact's public engineering account describes an asynchronous media pipeline: parallel transcription and shot detection, a join before scene assembly, per-scene extraction with neighbouring context, and one denormalized search document per scene. It does not disclose the model services. Their “proprietary large video context model” is a claim, not evidence that they trained base weights. The hard-to-copy part may be reliable orchestration, time alignment, grounding and editorial quality.

Stage to build	Practical starting point	Proof required
Audio	Clear proxy audio; compare existing subtitles, self-hosted speech recognition and managed Indic speech recognition.	Bengali word/character error and cue timing on labelled clips.
Scenes	Shot boundaries plus transcript/silence boundaries; sparse frames and constrained scene descriptions with evidence IDs.	Scene coverage, factuality and visual lift over an audio-only baseline.
Metadata	Hierarchical synopsis; map to Hoichoi-owned taxonomy IDs with an abstain option and curated Bengali labels.	Precision/recall and editor acceptance per field.
Serving	Versioned raw artifacts, field-level CMS review, approved title and scene indexes in Typesense.	Search relevance and p95 latency in three query forms.

Use durable job records and separate CPU media work from GPU inference. Start transcription as soon as audio is ready; assemble scenes only when transcription and shot detection have terminal results. Make every stage idempotent and keep source frames and transcript spans behind each claim. A service-scoped clear proxy avoids sending 5–13 GB masters through every analysis attempt; a 2 Mb/s proxy is about 0.9 GB per media hour before overhead. The full implementation and benchmark plan is recorded in the private research package.

clear analysis proxy + subtitle files → repeatable ingest job ↓ Bengali speech recognition + alignment · shot/frame sampling · optional faces/OCR ↓ evidence-linked scenes, synopsis, moments and time ranges ↓ our fixed taxonomy + bilingual labels → CMS field-level pending review ↓ approved title data → Typesense → browse/search · approved markers → players

Use what Hoichoi already runs

Mongo CMS remains the editorial authority. Store raw analysis and every suggestion version separately; accept or reject each field with an audit trail. The existing Typesense sync can index approved text and facet IDs. Start with title search, then add a scene collection keyed by title and timecode.

Work that still has to be built

Clear low-resolution proxies; a durable analysis queue; Bengali speech and visual evaluation; evidence-backed structured extraction; taxonomy mapping with abstention; per-field approval; error/retry and cost accounting; a distinct skip-credits control plus KMP mapping fixes. Model and vector-engine choices require a benchmark.

Reproduce the vendor's output shape with an artifact version, media checksum, model/prompt versions, resource state, time ranges, source evidence and approved taxonomy IDs. Do not publish face identity, captions, warnings or skip markers without editorial checks. Compare Tessact with a small in-house baseline on the same 12–15 diverse titles using Bengali word/character error, taxonomy precision and recall, hallucination rate, marker timing, latency and cost per media hour.

What remains unresolved

Tessact has not provided the direct Enterprise MSA/DPA, no-training terms, subprocessor list, derivative storage/region map, supported service API, rate limits, complete export scope, read-only S3 role flow, webhook delivery guarantee, Content Profile behaviour or 800-title throughput and price. The frontend shows Next.js and a Django API; public material points to a FastAPI AI service, AWS Mumbai services, a GCP suggestion service and S3, but does not identify the private model weights, vector engine, database or queue. Request those artefacts rather than treating a browser observation as proof.

11What the API shapes reveal

The private client responses expose useful boundaries, though they do not reveal the code or model behind scene extraction.

Observed signature	Grounded design choice
Asset detail has separate `phoenix_job`, `content_intelligence` and search-index status. A completed asset reported 259 shots, 217 dialogue lines and 1,566 evidence facets.	Track analysis, each output resource and search readiness separately; keep a durable job manifest.
Most resources share `content_intelligence_v1`, `artifact_version`, `state`, `reason` and a page cursor. Cursors bind to artifact, filters and scope and expire.	Use immutable versioned artifacts; drain pages promptly and validate expected counts. HTTP 200 alone never means complete data.
The face route uses a different `face_people_v1` job and offset pagination. Most groups are unknown; audio returned 409.	Keep face clustering optional and separate from canonical cast and speaker identity.
Phoenix search returns rewritten query, 478 candidates in one sample, a 200-result rank window, rerank fields and richer flattened scene fields than the JSON scene export.	Build our own approved title and scene projections; benchmark candidate retrieval and reranking. Do not assume all search documents are exportable.
Seven search suggestions label their composer `gemini_vertex` and model `gemini-2.5-pro`. A display-title model label names `gemini-2.5-flash-lite` while that feature is disabled.	Existing models are a plausible component choice. These labels establish helper-path configuration, not the Content Intelligence scene model.

The public AI-service schema lists chat, metadata autofill, captions and remix routes, but not the Content Intelligence resource contract. An authenticated request for the Django schema returned HTTP 500. The private research package holds the complete response-shape atlas and the raw page export. These internal routes remain undocumented for partner use.

12Complete export evidence

Open an asset, then a resource, to inspect the full saved JSON and its exact observed field structure. All 267 saved data files, including the 141 response pages, search response and lifecycle event logs, are indexed in the manifest and ZIP. This page loads each JSON only when opened.

Custody and exactness

The downloadable document ZIP preserves exported values except 1111 expiring signed URL values, whose query strings are marked REDACTED_EXPIRING_SIGNATURE. The byte-exact ZIP is owner-only in the local research workspace and is intentionally absent from the site staging folder. Neither archive includes media, login state, upload URLs or HARs. The merged JSON is a client-combined view; original response pages are in the archive.

Download the complete sanitized export ZIP · Per-file SHA-256 manifest. The archive includes the API signature atlas, build/evaluation plan, assessment and live findings. The controlled run ledger stays owner-only.

How to trace a claim

Claim in this report	Primary saved evidence	Interpretation
Resource fields, states and counts	`survey/*__resource__page_N.json`; each asset/resource detail below links its merged view	`references/API-SIGNATURES-AND-STACK.md`
Search candidates, rank window, result fields and measured latency	Saved Phoenix response	Search section and API atlas; one request is not a p95 distribution
Upload, search, generic index, face and CI timing	Asset/event log and face-job log	Owner-only run ledger has timestamp arithmetic and poll bounds
Same-media output differences	Comparison record plus old/new resource JSON	Snapshot variation, with one-week version drift uncontrolled
Build choices and evaluation targets	Observed exports, `references/internal-stack-map.md` and `references/public-research.md`	`references/VIDEO-INTELLIGENCE-BUILD.md`

External use and open source comparison

The public pages describe CSV/XLSX custom-metadata export, transcript downloads, scene MP4 export and an “analyzing complete” webhook carrying a metadata URL. They do not publish a supported scene-level JSON/search API, an SDK or a bulk embedding/index export. The saved logged-in client contract let us extract the 13 paths below, subject to pagination and unavailable states. The full public-source matrix and PoC method are in references/PUBLIC-INTEGRATION-AND-OSS-POC.md inside the ZIP.

A local FFmpeg scene-score pass on the controlled video found 208 cut candidates at threshold 0.15 in 24 seconds, versus Tessact’s 213 reported shots. Tessact shot timestamps were not exportable, so boundary accuracy is unmeasured. Tesseract OCR found late credit-like text; a 97-scene SQLite FTS5 index placed the same saved scene first for one query, but it indexed Tessact-derived descriptions. ASR and vision-language model checkpoints were unavailable locally and were not benchmarked.

The full candidate register, output-scope matrix, reference-label plan, stage validators and comparison protocol are in references/OSS-OPTIONS-AND-VALIDATION.md inside the ZIP. Tessact is a comparator; independent labels decide quality.

Local versus API feasibility, prerequisite checkpoints, reviewer effort and the per-stage cost worksheet are in references/POC-REQUIREMENTS-AND-COST.md. Dated list rates, Bengali support, break-even arithmetic and the API/EC2/hybrid recommendation are in references/EC2-VS-API-PRICING.md with references/pilot-cost-model.py. Model throughput, quality and Tessact unit billing remain unmeasured.

Observed response envelopes

Surface	Saved contract	Readiness and pagination
Content Intelligence resources	`schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page`	`available / partial / unavailable`; artifact-bound cursor on multi-page resources.
People / face	`schema_version: face_people_v1`, job, groups, suppression count, source revision	Separate job, offset pages; audio request returned 409.
Editable documents	Document/revision IDs, source and artifact versions, content and original content	Distinct editable projection; no edit propagation was probed.
Library/search	Captured response fields and route signatures in the API atlas	Search readiness, generic index completion and CI publication were separate in the controlled run.

Search response and lifecycle evidence

Exact saved Phoenix search response · query, candidates, rank window, results and latency

Observed structure: top: query, effective_query, intent, query_mode, search_mode, runtime_mode, match_by, cache_hit, cache_status, result_count, total_candidates, page_size, rank_window, total_ranked_results, has_more, group_key, display_title_status, display_title_model, applied_filters, results, search_request_id, search_session_id, search_trace_id, latency_ms_total

Saved response JSON

Open this section to load the full JSON.

Controlled-run measurements: source media manifest, original asset detail, resource comparison, face output, upload and CI event log, face job log, and active jobs observations. Event logs are newline-delimited JSON and are included in the ZIP.

All captured assets and resources

Dotta EP01.mp3 · survey · 13 resource slots

Saved asset ID: 79c3d834-e216-492a-a40e-73a124d38547. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: capabilities, capability_profile, composition_preview, counts, duration_sec, genres, languages, moods, overview, representative_frame_ref, structure_summary, themes, top_moments

Merged export: Dotta_EP01_mp3__summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 5 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items; item: id, mood, title, genres, number, end_sec, setting, location, synopsis, frame_ref, packaging, start_sec, shot_count, time_of_day, duration_sec, evidence_refs, indoor_outdoor, people_evidence, narrative_function

Merged export: Dotta_EP01_mp3__scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 4 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items; item: id, title, number, end_sec, purpose, synopsis, frame_ref, start_sec, duration_sec, evidence_refs, related_ranges, related_shot_ids, related_scene_ids

Merged export: Dotta_EP01_mp3__segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 5 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items; item: id, beat, title, scores, end_sec, frame_ref, start_sec, description, duration_sec, event_end_sec, evidence_refs, why_important, event_start_sec, score_semantics

Merged export: Dotta_EP01_mp3__moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 4 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items; item: score, end_sec, start_sec, dialogue_id, evidence_refs, speaker_label, text_original, dialogue_function

Merged export: Dotta_EP01_mp3__quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 149 items · 2 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, available_translation_languages; item: id, end_sec, start_sec, evidence_refs, speaker_label, text_original, is_exact_quote, text_normalized, transliteration, dialogue_function, quote_importance_score

Merged export: Dotta_EP01_mp3__dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 4 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items; item: id, label, support, category, field_sources, last_seen_sec, thumbnail_ref, first_seen_sec, density_summary, identity_status, occurrence_count

Merged export: Dotta_EP01_mp3__settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people — no saved JSON; audio face route returned HTTP 409.

timeline · available · — items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: collection_info, composition_segments, duration_sec, editorial_signals, moment_markers, resolution, series, series_capabilities, story_beats

Merged export: Dotta_EP01_mp3__timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, limit, total_languages, truncated, jobs

Merged export: Dotta_EP01_mp3__translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items

Merged export: Dotta_EP01_mp3__signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Observed structure: top: document_id, document_kind, language, source_version, artifact_version, translation_version, revision_id, version, can_edit, content, original_content, updated_at, updated_by, search_sync, previous_source_available, previous_sources, _export; content: detailed_synopsis, contextual_synopsis

Merged export: Dotta_EP01_mp3__documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: Dotta_EP01_mp3__documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

hoi_byomkesh_bng_series01_season01_episode01.mp4 · survey · 13 resource slots

Saved asset ID: 017fc19f-576d-47c4-9d63-baa89e0f112a. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 25 items · 2 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 21 items · 2 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 12 items · 1 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 12 items · 1 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 421 items · 5 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 126 items · 3 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people · completed · 175 items · 4 page(s)

Observed structure: top: schema_version, asset_id, state, job, data, page, _export; data: groups, total, suppressed_total, reviewable_only; item: group_id, asset_id, status, person_id, display_name, suggested_person_id, suggested_display_name, match_confidence, representative_crop_url, appearance_count, screen_time_sec, first_seen_sec, last_seen_sec, media_type, asset_duration_sec, source_revision, ranges, created_at, updated_at

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__people.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

timeline · available · — items · 1 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, limit, total_languages, truncated, jobs

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 3 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items; item: id, kind, label, source, end_sec, start_sec, signal_type, evidence_refs, signal_semantics, is_editorial_exclusion

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: hoi_byomkesh_bng_series01_season01_episode01_mp4__documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

hoi_indubala_bhaater_hotel_bng_series01_season01_episode01.mp4 · survey · 13 resource slots

Saved asset ID: 26842d35-2fd7-4474-b026-c117762f38c5. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 15 items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 5 items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 4 items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 6 items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 217 items · 3 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 72 items · 2 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people · completed · 106 items · 3 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__people.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

timeline · available · — items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, limit, total_languages, truncated, jobs

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 19 items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: hoi_indubala_bhaater_hotel_bng_series01_season01_episode01_mp4__documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

jamai_bodol.mp4 · survey · 13 resource slots

Saved asset ID: 4b49a4be-865d-411b-b036-00f394da5ea7. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Merged export: jamai_bodol_mp4__summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 12 items · 1 page(s)

Merged export: jamai_bodol_mp4__scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 7 items · 1 page(s)

Merged export: jamai_bodol_mp4__segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 4 items · 1 page(s)

Merged export: jamai_bodol_mp4__moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 5 items · 1 page(s)

Merged export: jamai_bodol_mp4__quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 139 items · 2 page(s)

Merged export: jamai_bodol_mp4__dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 51 items · 2 page(s)

Merged export: jamai_bodol_mp4__settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people · completed · 21 items · 1 page(s)

Merged export: jamai_bodol_mp4__people.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

timeline · available · — items · 1 page(s)

Merged export: jamai_bodol_mp4__timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, limit, total_languages, truncated, jobs

Merged export: jamai_bodol_mp4__translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 1 items · 1 page(s)

Merged export: jamai_bodol_mp4__signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Merged export: jamai_bodol_mp4__documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: jamai_bodol_mp4__documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

meye_bela_ep04.mp4 · survey · 13 resource slots

Saved asset ID: 98e6457d-c752-4e7e-99a6-aa842d092661. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Merged export: meye_bela_ep04_mp4__summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 10 items · 1 page(s)

Merged export: meye_bela_ep04_mp4__scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 8 items · 1 page(s)

Merged export: meye_bela_ep04_mp4__segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 6 items · 1 page(s)

Merged export: meye_bela_ep04_mp4__moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 5 items · 1 page(s)

Merged export: meye_bela_ep04_mp4__quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 121 items · 2 page(s)

Merged export: meye_bela_ep04_mp4__dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 49 items · 1 page(s)

Merged export: meye_bela_ep04_mp4__settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people · completed · 40 items · 1 page(s)

Merged export: meye_bela_ep04_mp4__people.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

timeline · available · — items · 1 page(s)

Merged export: meye_bela_ep04_mp4__timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, limit, total_languages, truncated, jobs

Merged export: meye_bela_ep04_mp4__translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 16 items · 1 page(s)

Merged export: meye_bela_ep04_mp4__signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Merged export: meye_bela_ep04_mp4__documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: meye_bela_ep04_mp4__documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

meye_bela_ep03.mp4 · survey · 13 resource slots

Saved asset ID: f44de0be-9bf6-4fcc-ac9a-178e9f6e9dc5. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Merged export: meye_bela_ep03_mp4__summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 13 items · 1 page(s)

Merged export: meye_bela_ep03_mp4__scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 9 items · 1 page(s)

Merged export: meye_bela_ep03_mp4__segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 5 items · 1 page(s)

Merged export: meye_bela_ep03_mp4__moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 4 items · 1 page(s)

Merged export: meye_bela_ep03_mp4__quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 156 items · 2 page(s)

Merged export: meye_bela_ep03_mp4__dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 40 items · 1 page(s)

Merged export: meye_bela_ep03_mp4__settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people · completed · 46 items · 1 page(s)

Merged export: meye_bela_ep03_mp4__people.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

timeline · available · — items · 1 page(s)

Merged export: meye_bela_ep03_mp4__timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, limit, total_languages, truncated, jobs

Merged export: meye_bela_ep03_mp4__translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 4 items · 1 page(s)

Merged export: meye_bela_ep03_mp4__signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Merged export: meye_bela_ep03_mp4__documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: meye_bela_ep03_mp4__documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

meye_bela_ep02.mp4 · survey · 13 resource slots

Saved asset ID: 3b849bf8-b08c-4ba4-bac9-dc4d5a06b20b. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Merged export: meye_bela_ep02_mp4__summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 11 items · 1 page(s)

Merged export: meye_bela_ep02_mp4__scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 8 items · 1 page(s)

Merged export: meye_bela_ep02_mp4__segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 5 items · 1 page(s)

Merged export: meye_bela_ep02_mp4__moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 5 items · 1 page(s)

Merged export: meye_bela_ep02_mp4__quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 111 items · 2 page(s)

Merged export: meye_bela_ep02_mp4__dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 56 items · 2 page(s)

Merged export: meye_bela_ep02_mp4__settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people · completed · 36 items · 1 page(s)

Merged export: meye_bela_ep02_mp4__people.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

timeline · available · — items · 1 page(s)

Merged export: meye_bela_ep02_mp4__timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, limit, total_languages, truncated, jobs

Merged export: meye_bela_ep02_mp4__translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 5 items · 1 page(s)

Merged export: meye_bela_ep02_mp4__signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Merged export: meye_bela_ep02_mp4__documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: meye_bela_ep02_mp4__documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

meye_bela_ep01.mp4 · survey · 13 resource slots

Saved asset ID: 17193e02-5eb2-4148-99c7-c833575fe6e6. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Merged export: meye_bela_ep01_mp4__summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 6 items · 1 page(s)

Merged export: meye_bela_ep01_mp4__scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 4 items · 1 page(s)

Merged export: meye_bela_ep01_mp4__segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 3 items · 1 page(s)

Merged export: meye_bela_ep01_mp4__moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 5 items · 1 page(s)

Merged export: meye_bela_ep01_mp4__quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 92 items · 1 page(s)

Merged export: meye_bela_ep01_mp4__dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 41 items · 1 page(s)

Merged export: meye_bela_ep01_mp4__settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people · completed · 49 items · 2 page(s)

Merged export: meye_bela_ep01_mp4__people.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

timeline · available · — items · 1 page(s)

Merged export: meye_bela_ep01_mp4__timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page, _export; data: items, limit, total_languages, truncated, jobs

Merged export: meye_bela_ep01_mp4__translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 5 items · 1 page(s)

Merged export: meye_bela_ep01_mp4__signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Merged export: meye_bela_ep01_mp4__documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: meye_bela_ep01_mp4__documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

Test Object.mp4 · controlled · 13 resource slots

Saved asset ID: bce3b81a-3eed-4fe0-9c52-1890544a6fe3. Each resource below opens its complete merged JSON. Use the ZIP for every original response page.

summary · available · — items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: capabilities, capability_profile, composition_preview, counts, duration_sec, genres, languages, moods, overview, representative_frame_ref, structure_summary, themes, top_moments

Merged export: summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

scenes · available · 8 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: items; item: id, mood, title, genres, number, end_sec, setting, location, synopsis, frame_ref, packaging, start_sec, shot_count, time_of_day, duration_sec, evidence_refs, indoor_outdoor, people_evidence, narrative_function

Merged export: scenes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

segments · available · 4 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: items; item: id, title, number, end_sec, purpose, synopsis, frame_ref, start_sec, duration_sec, evidence_refs, related_ranges, related_shot_ids, related_scene_ids

Merged export: segments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

moments · available · 3 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: items; item: id, beat, title, scores, end_sec, frame_ref, start_sec, description, duration_sec, event_end_sec, evidence_refs, why_important, event_start_sec, score_semantics

Merged export: moments.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

quotes · available · 5 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: items; item: score, end_sec, start_sec, dialogue_id, evidence_refs, speaker_label, text_original, dialogue_function

Merged export: quotes.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

dialogue · available · 94 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: items, available_translation_languages; item: id, end_sec, start_sec, evidence_refs, speaker_label, text_original, is_exact_quote, text_normalized, transliteration, dialogue_function, quote_importance_score

Merged export: dialogue.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

settings · available · 32 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: items; item: id, label, support, category, field_sources, last_seen_sec, thumbnail_ref, first_seen_sec, density_summary, identity_status, occurrence_count

Merged export: settings.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

people · completed · 49 items · 2 page(s)

Observed structure: top: schema_version, asset_id, state, job, data, page; data: groups, total, suppressed_total, reviewable_only; item: group_id, asset_id, status, person_id, display_name, suggested_person_id, suggested_display_name, match_confidence, representative_crop_url, appearance_count, screen_time_sec, first_seen_sec, last_seen_sec, media_type, asset_duration_sec, source_revision, ranges, created_at, updated_at

Merged export: people.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

timeline · available · — items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: collection_info, composition_segments, duration_sec, editorial_signals, moment_markers, resolution, series, series_capabilities, story_beats

Merged export: timeline.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

translations · unavailable · 0 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: items, limit, total_languages, truncated, jobs

Merged export: translations.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

signals · partial · 5 items · 1 page(s)

Observed structure: top: schema_version, resource, asset_id, artifact_version, generated_at, state, reason, data, page; data: items; item: id, kind, label, source, end_sec, start_sec, signal_type, evidence_refs, signal_semantics, is_editorial_exclusion

Merged export: signals.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_summary · document · — items · 1 page(s)

Merged export: documents_summary.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

documents_subtitles_original · document · — items · 1 page(s)

Merged export: documents_subtitles_original.json. Response pages are indexed in manifest.json and the ZIP.

Open this section to load the full JSON.

Findings verified 2026-09-28 against the live Tessact workspace (org b97887e9-4ae5-4999-834a-fe8f692111ea, workspace 0c93e618-591d-48ce-94a2-348968a72ebf), 8 sample titles, read-only capture + GET export. Full evidence, the endpoint map, and the export client are held internally. Public-source research and the internal stack map are separate documents.