Vitra Translate Suite

Create a new brand kit by scraping a brand's homepage via POST https://api.translate.photo/v1/service/image-creator/brand-kit (Content-Type: application/json). JSON body: {url: '<homepage-url>'}. Returns the new brand kit including its brand_kit_id. Call this ONLY when Vitra_ListBrandKits surfaced no kit matching the Steward's brand/domain. The scrape, its output, and the resulting brand_kit_id are INTERNAL — never shown to the Steward; the only Steward-facing prompt is asking for the website URL when one is needed and absent from context. Runtime mapping (Mind passes snake_case; runtime constructs the API JSON body): url <- website_url. Hard reject any website_url whose scheme is not http(s). Errors arrive as standard non-2xx HTTP (401/403 = auth rejected; any other non-2xx = request/server error).

Submit an image for text translation via POST https://api.translate.photo/v1/image-translator/direct-translate. JSON body: {fileUrl: '<source-url>', sourceLanguage: '<value-from-Vitra_ListImageLanguages>', targetLanguage: '<value-from-Vitra_ListImageLanguages>'}. The vendor fetches the fileUrl server-side and enforces the 10MB cap — no client-side multipart upload, no client-side download. Async — returns {jobId, translationVersionId} immediately; you MUST call Vitra_GetTranslationStatus to retrieve the translatedImageUrl. NEVER pass ISO codes ('es', 'fr', 'hi') — pass the identifier the API returned in the 'value' field of Vitra_ListImageLanguages (e.g. 'spanish', 'french', 'hindi'). Auto-detect aliases ('auto-detect', 'Auto Detect', 'autodetect', 'detect') are normalized to the literal 'auto' before send. This endpoint's errors arrive as standard HTTP status (401/403 = auth rejected; any other non-2xx = request/server error — read the JSON body for the message). Runtime mapping: source_url ? fileUrl field; source_language ? sourceLanguage field; target_language ? targetLanguage field. Hard reject any source_url whose scheme is not http(s).

Edit a previously generated creative via POST https://api.translate.photo/v1/service/image-creator/image/edit (Content-Type: application/json). JSON body: {assetId: '<asset_id>', instruction: '<plain-language edit>'}. SYNCHRONOUS — returns a fresh asset ({status: 'ok', data: {asset_id, asset_url, gallery_url}}) directly; no poll. asset_id MUST be one captured from a Vitra_GenerateImage (or a prior Vitra_EditImage / Vitra_TranslateImageLayout) response IN THE CURRENT SESSION — never Steward-supplied or fabricated. The returned asset_id SUPERSEDES the prior one for further follow-ups. Present ONLY the new asset_url + gallery_url; NEVER show asset_id. Runtime mapping (Mind passes snake_case; runtime constructs the API JSON body): assetId <- asset_id; instruction <- instruction. Errors arrive as standard non-2xx HTTP (401/403 = auth rejected; any other non-2xx = request/server error).

Generate a structured, brand-kit-aware creative image via POST https://api.translate.photo/v1/service/image-creator/image/generate (Content-Type: application/json). JSON body: {brandKit: '<brand_kit_id>', channel: '<platform-format>', headline?: '<copy>', cta?: '<copy>', enrich?: <bool>}. SYNCHRONOUS — on an HTTP 2xx the response carries the finished asset (per the reference SKILL.md the success shape is {status: 'ok', data: {asset_id, asset_url, gallery_url}}); there is NO async job and NO poll tool. Capture asset_id as INTERNAL state (thread it into Vitra_EditImage / Vitra_TranslateImageLayout) and present ONLY asset_url (direct image link) and gallery_url to the Steward — NEVER asset_id, NEVER brand_kit_id. brand_kit_id must be a resolved id from Vitra_ListBrandKits or Vitra_CreateBrandKit (not Steward-supplied). channel is a platform-format identifier (e.g. 'instagram_post', 'facebook_cover'). Runtime mapping (Mind passes snake_case; runtime constructs the API JSON body): brandKit <- brand_kit_id; channel <- channel; headline <- headline; cta <- cta; enrich <- enrich. Errors arrive as standard non-2xx HTTP (401/403 = auth rejected; any other non-2xx = request/server error).

Poll the async dubbing result via GET https://api.translate.video/v1/service/status/{processId}. Shared between Vitra_SubmitAudioDub and Vitra_SubmitVideoDub (same processId namespace, same status endpoint). Returns {status, data: [...], ...}. DONE statuses (uppercase): 'DONE', 'COMPLETED', 'SUCCESS'. FAILED statuses (uppercase): 'FAILED', 'ERROR', 'CANCELLED'. On a completed status, the output URL is at data[0].url — extract that and present it to the Steward. Recommended polling cadence: every 10 seconds within a 600-second total budget (dubbing is slower than image translation). NEVER expose processId to the Steward as a result — it is an intermediate identifier. Error handling: callers branch on raw HTTP status — a non-200 response is an error, and 401/403 means the va-token was rejected. On a completed status with empty data or no data[0].url, treat it as an error (surface it and stop; do not loop). Runtime mapping: the JSON Schema parameter 'process_id' (snake_case for Mind ergonomics) is substituted into the URL path segment '{processId}' (camelCase per Vitra's API). No request body.

Poll the async image-translation result via GET https://api.translate.photo/v1/image-translator/{jobId}/translations/{translationVersionId}. Returns {status, translatedImageUrl}. Status values include 'pending', 'translating', 'completed' (also 'complete', 'done', 'success', 'succeeded', 'finished' — case-insensitive), and failure states ('failed', 'error', 'errored', 'cancelled', 'canceled'). translatedImageUrl is the final cloud URL the Mind returns to the Steward — only present when status is in the completed family AND the URL starts with http(s). Recommended polling cadence: every 5 seconds within a 300-second total budget. NEVER expose jobId or translationVersionId to the Steward as a result — they are intermediate identifiers. This endpoint's errors arrive as standard HTTP status (401/403 = auth rejected). Runtime mapping: the JSON Schema parameter 'job_id' (snake_case for Mind ergonomics) is substituted into the URL path segment '{jobId}' (camelCase per Vitra's API); 'translation_version_id' is substituted into '{translationVersionId}'. No request body.

List languages supported by the vitra.ai audio dubbing service via GET https://api.translate.video/v1/service/audio/languages/list. Returns {data: [{value, label, ...}]} where 'value' is the locale-form identifier the dubbing endpoint accepts (e.g. 'hindi_india', 'spanish_spain', 'french_france') and 'label' is the human-readable name. Use this BEFORE Vitra_SubmitAudioDub whenever you are unsure of the exact target language locale or when the Steward gives a plain language name ('Spanish') that needs disambiguation to a locale ('spanish_spain' vs 'spanish_mexico'). Response is cacheable for ~1 hour. NOTE: this endpoint is called WITHOUT X-Client-Source: agent — errors arrive as standard non-2xx HTTP.

List the brand kits already configured for this Vitra account via GET https://api.translate.photo/v1/service/image-creator/brand-kits. Returns the brand kits, each with a brand_kit_id and brand/page/domain metadata. Use this BEFORE Vitra_GenerateImage to SILENTLY resolve which brand kit to use: match the Steward's requested brand / page / website domain to an existing kit and reuse its brand_kit_id. CRITICAL — the brand-kit list, kit details, and brand_kit_id are INTERNAL: never show them to the Steward. If no kit matches and a homepage URL is available, scrape a new one via Vitra_CreateBrandKit. Response is cacheable for ~1 hour (LTM 'vitra_brand_kit_list'). This endpoint is on the image-creator product line (api.translate.photo /v1/service/image-creator), distinct from the image-text-translation endpoints. Errors arrive as standard non-2xx HTTP (401/403 = auth rejected; any other non-2xx = request/server error with a JSON payload). No request body.

List languages supported by the vitra.ai image translator via GET https://api.translate.photo/v1/language/list. Returns an array of language objects, each with a 'value' field (the identifier the submit endpoint accepts, e.g. 'spanish') and a 'name' (or 'label') field (human-readable). The script normalizer also accepts 'code'/'key'/'language'/'id' as fallback identifier fields. Use this BEFORE Vitra_DirectTranslate whenever you are unsure of the exact target/source language identifier or when the Steward gives an ISO code ('es', 'fr') that needs to be mapped. Response is cacheable for ~1 hour. NOTE: this endpoint is called WITHOUT X-Client-Source: agent — errors arrive as standard non-2xx HTTP (401/403/4xx with payload), not as HTTP 200 + success:false.

List source languages supported by the vitra.ai transcription (speech-to-text) service via GET https://api.translate.video/v1/service/transcription/languages. Returns {data: [{value, label}]} where 'value' is the locale-form identifier the transcription submit endpoint accepts (e.g. 'english_united_states', 'english_united_kingdom', 'hindi_india', 'spanish_spain') and 'label' is the human-readable name. Use this BEFORE Vitra_SubmitTranscription whenever you are unsure of the exact source locale or when the Steward gives a plain language name ('English') that needs disambiguation between locales ('english_united_states' vs 'english_united_kingdom'). Response is cacheable for ~1 hour. NOTE: this endpoint is called WITHOUT X-Client-Source: agent — errors arrive as standard non-2xx HTTP.

List languages supported by the vitra.ai video dubbing service via GET https://api.translate.video/v1/service/video/languages/list. Returns {data: [{value, label, ...}]} where 'value' is the locale-form identifier the dubbing endpoint accepts (e.g. 'hindi_india', 'spanish_spain') and 'label' is the human-readable name. Use this BEFORE Vitra_SubmitVideoDub whenever you are unsure of the exact target language locale. Audio and video share the same locale convention but are listed under different endpoints — call this for video, Vitra_ListAudioLanguages for audio. Response is cacheable for ~1 hour. NOTE: this endpoint is called WITHOUT X-Client-Source: agent — errors arrive as standard non-2xx HTTP.

Submit an audio file for dubbing via POST https://api.translate.video/v1/service/dubbing (Content-Type: application/json). JSON body: {processName: 'Audio Dubbing: <filename>', sourceLanguage: 'auto' | identifier, targetLanguages: [{language: '<value-from-Vitra_ListAudioLanguages>'}], url: '<source-url>'}. Async — returns {processId, ...}; you MUST call Vitra_GetDubbingStatus to retrieve the dubbed audio URL. The vitra.ai source-language default is 'auto'; do NOT ask the Steward for source language unless they volunteer it. target_language MUST be the 'value' field of a Vitra_ListAudioLanguages entry (locale-form: 'hindi_india', not 'hindi'). http(s) cloud URLs only (https strongly preferred). This endpoint's errors arrive as standard non-2xx HTTP: 401/403 = auth rejected, any other non-2xx = request/server error (read the JSON body for the message). Runtime mapping (Mind passes snake_case params; runtime constructs the API JSON body): url ? source_url; sourceLanguage ? source_language; targetLanguages ? [{language: target_language}] (wrap the single target into a one-element array of objects); processName ? 'Audio Dubbing: ' + the source URL's filename (the basename of the source URL, matching the reference script's Path(url).name). Hard reject any source_url whose scheme is not http(s).

Submit an audio or video file for transcription (speech-to-text) via POST https://api.translate.video/v1/service/transcription (Content-Type: application/json). JSON body: {processName: 'Transcription: <filename>', sourceLanguage: '<value-from-Vitra_ListTranscriptionLanguages>', url: '<source-url>'}. Async — returns {processId} (HTTP 200 OR 201 are both success); you MUST call Vitra_GetDubbingStatus to retrieve the transcripts (the shared status endpoint /v1/service/status/{processId} also serves audio + video dubbing; the response's `data` field is the transcript array [{st, et, text}, ...] for transcription, NOT a single dubbed-output URL like dubbing's data[0].url). NOTE: source_language is REQUIRED — transcription does NOT accept 'auto' (unlike audio/video dubbing). Pass a locale-form 'value' from Vitra_ListTranscriptionLanguages (e.g. 'english_united_states', 'hindi_india'); plain names ('english', 'hindi') and ISO codes ('en', 'hi') are rejected. http(s) cloud URLs only (https strongly preferred). This endpoint's errors arrive as standard non-2xx HTTP: 401/403 = auth rejected, any other non-2xx = request/server error (read the JSON body for the message). Runtime mapping (Mind passes snake_case params; runtime constructs the API JSON body): url ? source_url; sourceLanguage ? source_language; processName ? 'Transcription: ' + the source URL's filename (the basename, matching the reference script's Path(url).name). Hard reject any source_url whose scheme is not http(s).

Submit a video file for dubbing via POST https://api.translate.video/v1/service/dubbing (Content-Type: application/json). JSON body: {processName: 'Dubbing: <filename>', sourceLanguage: 'auto' | identifier, targetLanguages: [{language: '<value-from-Vitra_ListVideoLanguages>'}], url: '<source-url>'}. Async — returns {processId, ...}; you MUST call Vitra_GetDubbingStatus to retrieve the dubbed video URL. The vitra.ai source-language default is 'auto'; do NOT ask the Steward for source language unless they volunteer it. target_language MUST be the 'value' field of a Vitra_ListVideoLanguages entry (locale-form). http(s) cloud URLs only (https strongly preferred). This endpoint's errors arrive as standard non-2xx HTTP: 401/403 = auth rejected, any other non-2xx = request/server error (read the JSON body for the message). Runtime mapping (Mind passes snake_case params; runtime constructs the API JSON body): url ? source_url; sourceLanguage ? source_language; targetLanguages ? [{language: target_language}] (wrap the single target into a one-element array of objects); processName ? 'Dubbing: ' + the source URL's filename (the basename of the source URL, matching the reference script's Path(url).name). Hard reject any source_url whose scheme is not http(s).

Re-render the copy of a previously generated creative in another language via POST https://api.translate.photo/v1/service/image-creator/image/translate (Content-Type: application/json). JSON body: {assetId: '<asset_id>', locale: '<bcp47-iso-code>'}. SYNCHRONOUS — returns a fresh asset ({status: 'ok', data: {asset_id, asset_url, gallery_url}}) directly; no poll. CRITICAL DISTINCTION: `locale` here is a BCP47 ISO code ('es', 'fr', 'hi', 'de', 'ja', 'pt-br') — the OPPOSITE of Vitra_DirectTranslate, which rejects ISO codes and requires a Vitra_ListImageLanguages identifier ('spanish'). This tool needs an asset_id (a layout generated in THIS skill) and NO source URL; Vitra_DirectTranslate needs a standalone image URL and NO asset_id. asset_id MUST be captured from a generate/edit/translate response in the current session. The returned asset_id supersedes the prior one. Present ONLY the new asset_url + gallery_url; NEVER show asset_id. Runtime mapping (Mind passes snake_case; runtime constructs the API JSON body): assetId <- asset_id; locale <- locale. Errors arrive as standard non-2xx HTTP (401/403 = auth rejected; any other non-2xx = request/server error).