Studio model IDs, provider endpoints, and display names

Continue from identification to a model-specific decision

Use the model directory to refresh the launch choices, then open a model guide only after the Studio ID and workflow are fixed. The linked comparison is appropriate when the decision is specifically fast image iteration versus a more detailed FLUX path; it should not replace the exact identifier record.

The current Studio ID, display-name, and endpoint map

OfflineCreator Studio's deployed catalog currently contains six launch entries. Each entry keeps three different identifiers side by side: a short Studio ID used by Studio and MCP calls, a human-facing display name, and a fal provider endpoint used by the service integration. The exact pairs are `flux-schnell` / FLUX Schnell / `fal-ai/flux-1/schnell`; `flux-pro` / FLUX 1.1 Pro Ultra / `fal-ai/flux-pro/v1.1-ultra`; and `recraft-v3` / Recraft V3 / `fal-ai/recraft/v3/text-to-image`.

The same catalog maps `kling-text` to Kling 2.6 Pro and `fal-ai/kling-video/v2.6/pro/text-to-video`; `kling-image` to Kling 2.6 Pro Motion and `fal-ai/kling-video/v2.6/pro/image-to-video`; and `veo-fast` to Veo 3.1 Fast and `fal-ai/veo3.1/fast`. The mapping also records workflow, so the two Kling display names are not interchangeable: one is text-to-video and the Motion entry is image-to-video.

FLUX Schnell

Studio ID: flux-schnellProvider endpoint: fal-ai/flux-1/schnell · Workflow: text-to-image

FLUX 1.1 Pro Ultra

Studio ID: flux-proProvider endpoint: fal-ai/flux-pro/v1.1-ultra · Workflow: text-to-image

Recraft V3

Studio ID: recraft-v3Provider endpoint: fal-ai/recraft/v3/text-to-image · Workflow: text-to-image

Kling 2.6 Pro

Studio ID: kling-textProvider endpoint: fal-ai/kling-video/v2.6/pro/text-to-video · Workflow: text-to-video

Kling 2.6 Pro Motion

Studio ID: kling-imageProvider endpoint: fal-ai/kling-video/v2.6/pro/image-to-video · Workflow: image-to-video

Veo 3.1 Fast

Studio ID: veo-fastProvider endpoint: fal-ai/veo3.1/fast · Workflow: text-to-video

Keep the three identifier namespaces separate

A display name is for recognition, not request routing. “Kling 2.6 Pro” identifies a model family in prose, but Studio currently has separate text-to-video and image-to-video entries whose names differ only by “Motion.” A command, saved manifest, or comparison table that records only “Kling” loses the workflow distinction and cannot be audited against the configured route.

A Studio ID and a fal endpoint are also different namespaces. The deployed catalog stores them in separate `id` and `endpoint` fields. The published MCP `generate` schema calls its request field `modelId` and describes it as the model ID returned by `list_models`, with `flux-schnell` as its example. fal's own documentation labels strings such as `fal-ai/flux-1/schnell` as fal model IDs and request endpoints. Those provider strings verify the underlying provider contract; they are not values to substitute for Studio's MCP ID.

Display name

Use in headings and human reviewExample: FLUX 1.1 Pro Ultra. Do not use it as an assumed request key.

Studio model ID

Use in generate.modelIdExample: flux-pro. Refresh it from list_models before an operational run.

Provider endpoint

Use to audit the underlying fal schemaExample: fal-ai/flux-pro/v1.1-ultra. Do not paste it into generate.modelId unless Studio explicitly returns it there.

Avoid the ambiguous-family-name failure mode

Suppose an approval note says “use Kling 2.6 Pro” and an automation later translates that phrase into a request. The phrase does not say whether the job starts from text or from an image. The current mapping requires `kling-text` for the text-to-video entry and `kling-image` for the image-to-video Motion entry. Recording the Studio ID, workflow, and provider endpoint together turns the approval into an auditable selection instead of a fuzzy family-name match.

Apply the same discipline when comparing versions. The display name FLUX 1.1 Pro Ultra maps to the short Studio ID `flux-pro`, while its current fal endpoint includes the more specific path `v1.1-ultra`. A comparison that records only “FLUX Pro” can silently outlive the configured version. Preserve the exact triplet in generation manifests, test notes, and bug reports, then refresh it before reproducing the result.

Evidence boundary and monthly refresh rule

This research pass inspected `src/lib/models.ts` and confirmed that it still contains the six identifier triplets shown here, but the working file now also contains additional model definitions that are absent from the deployed public catalog bundle. Those additional definitions are not treated as production availability evidence. The citation therefore points readers to the accessible deployed bundle, which currently exposes the six Studio IDs, display names, workflows, and provider endpoints listed on this page. The public model page independently confirms the same six display names, workflows, and credit costs, while the published MCP package establishes how callers obtain and use a Studio ID. The six fal pages independently confirm the provider endpoint strings and categories.

Recent community evidence was insufficient and is not used for any claim. The current last30days result returned 32 items, but none established an OfflineCreator Studio identifier mapping: the visible clusters were off-topic or thin. Digg was rate-limited with HTTP 429, Reddit coverage was partial, and X was unavailable because it was not configured; those lanes remain unknown rather than quiet. Polymarket, Techmeme, and YouTube completed with no results, while the other successful lanes produced no usable page-specific evidence. This pass did not use production credentials, call the authenticated models API, submit a provider request, or verify a completed generation. Refresh the deployed bundle, an authorized `list_models` response suitable for publication, the MCP package, and the fal pages before editorial review or after a catalog change.