← falCONTENT HISTORY

Update to fal

Snapshot Nov 14, 2026 · 00:10 UTC · version 1.3.2

Collection source: downloaded plugin package. These snapshots do not have a confirmed matching collection source. Differences in file lists alone do not establish changes to the package.

WHAT CHANGED

Payment or plan references changed

Skill instructions — Now says: Bundled guidance is fallback craft. For inputs, recommend_model's `input_contract` and live validation at submission are authoritative. See [the shared...

Observed in published text. Live prices and checkout terms have not been verified by this change.

Save fal and check future changes in Following.

Up to 3 plugins free · no card required.

All fields and original values.

Skill instructions

Before

Sep 30, 2026 · 22:46 UTC

Use this skill for curated endpoint choice. It is not a replacement for the MCP tools. After choosing an endpoint, execute through the fal.ai MCP tools in this plugin. Do not call the genmedia CLI from this MCP skill bundle. ## Routing R...

After

Nov 14, 2026 · 00:10 UTC

Bundled guidance is fallback craft. For inputs, recommend_model's `input_contract` and live validation at submission are authoritative. See [the shared production contract](../shared/production-contract.md). Use current task-relevant pub...

Compare saved observations

Download comparison JSON

SKILL.md text changes

− removed · + added

--- before
+++ after
@@ -9,121 +9,54 @@
 
 # fal.ai Model Routing
 
-Use this skill for curated endpoint choice. It is not a replacement for the
-MCP tools. After choosing an endpoint, execute through the fal.ai MCP tools in
-this plugin. Do not call the genmedia CLI from this MCP skill bundle.
-
-## Routing Rules
-
-1. If the user did not name a specific endpoint, call `recommend_model` first
-   for a current catalog-ranked recommendation.
-2. Use the curated endpoints in this skill to evaluate the recommendation and
-   pick the artifact-appropriate route. If the recommendation is generic but
-   the task is specialized, use `search_models` to verify a better endpoint.
-3. Use live MCP discovery when the role is not covered, an endpoint is missing
-   or deprecated, or the user asks to compare options.
-4. Inspect the selected endpoint schema before using custom fields.
-5. Check pricing when cost, batch size, video duration, 3D, or high-resolution
-   output matters.
-6. Do not invent endpoint IDs or schema fields.
-
-## Image Generation
-
-Generic/simple image generation:
-
-- `fal-ai/flux-2/klein/9b`: default when the user asks for a basic image and
-  does not name a model or require premium text rendering, brand fidelity, or
-  final commercial quality.
-
-Text-heavy image work:
-
-- `openai/gpt-image-2`: best default for readable text, posters, packaging,
-  diagrams, UI mockups, book covers, exact-copy layouts, and high-end stills.
-  Use high quality when the result is final.
-- `fal-ai/nano-banana-pro`: second choice when text is important but cost or
-  availability matters.
-
-Premium still images:
-
-- `openai/gpt-image-2`: commercial realism, editorial photography, product
-  scenes, character anchors, and high-quality concept art.
-- `fal-ai/nano-banana-pro`: strong styled output and product-friendly images.
-- `fal-ai/nano-banana-2`: cheaper strong alternative.
-
-Fast draft images:
-
-- `fal-ai/flux-2/klein/9b`: quick concepts, mood exploration, rough layouts.
-  Do not use as final commercial delivery unless the user asks for speed/cost.
-
-## Image Editing
-
-Use for product edits, background replacement, inpainting, relighting, cleanup,
-object changes, outfit changes, and reference-driven composition.
-
-1. `fal-ai/flux-2/klein/9b/edit`: default for simple/generic image edits when
-   the user does not name a model.
-2. `fal-ai/nano-banana-pro/edit`
-3. `openai/gpt-image-2/edit`
-4. `fal-ai/bytedance/seedream/v5/lite/edit`
-
-For product fidelity, prefer reference/edit workflows over text-only generation.
-
-## Video
-
-Highest quality video:
-
-- Text to video: `bytedance/seedance-2.0/text-to-video`
-- Image to video: `bytedance/seedance-2.0/image-to-video`
-- Reference to video: `bytedance/seedance-2.0/reference-to-video`
-
-Fast/lower-cost video:
-
-- Text to video: `xai/grok-imagine-video/text-to-video`
-- Image to video: `xai/grok-imagine-video/image-to-video`
-- Video edit: `xai/grok-imagine-video/edit-video`
-
-Multi-shot or control-heavy video:
-
-1. `bytedance/seedance-2.0/text-to-video`
-2. `bytedance/seedance-2.0/image-to-video`
-3. `bytedance/seedance-2.0/reference-to-video`
-4. `fal-ai/kling-video/v3/pro/text-to-video`
-5. `fal-ai/kling-video/v3/pro/image-to-video`
-6. `alibaba/happy-horse/text-to-video`
-7. `alibaba/happy-horse/image-to-video`
-
-Use Kling when multi-prompt, element controls, or stronger prompt structure are
-needed. Use Happy Horse for brief, direct camera-language prompts.
-
-## UGC And Talking Head
-
-- Portrait plus audio: `veed/fabric-1.0`
-- Portrait plus text: `veed/fabric-1.0/text`
-- Avatar with visual direction: `fal-ai/creatify/aurora`
-- Existing footage with new speech: `fal-ai/sync-lipsync/v2`
-- Product b-roll: `bytedance/seedance-2.0/image-to-video`
-- Fast b-roll draft: `xai/grok-imagine-video/image-to-video`
-
-## Audio
-
-Use live catalog discovery for current TTS, music, sound effect, transcription,
-and cleanup endpoints. Audio model availability changes quickly, so inspect
-schemas and pricing before execution.
-
-## 3D
-
-- Premium image-to-3D or character assets: use Meshy v6 class endpoints when
-  available and inspect rigging/animation fields.
-- Hunyuan/Tripo-style endpoints can be alternatives for faster drafts or
-  different geometry outputs.
-- Always run 3D jobs async and return model URLs plus any texture/material URLs.
-
-## Quality Guardrail
-
-Model choice is only valid if it matches the artifact role:
-
-- Text-heavy deliverables need text-capable image models or external layout.
-- Product deliverables need reference/edit workflows when packaging must hold.
-- Talking-head deliverables need audio/script length checks.
-- Story deliverables need continuity anchors across shots.
-- 3D deliverables need output format, scale, material, and rigging checks.
+Bundled guidance is fallback craft. For inputs, recommend_model's
+`input_contract` and live validation at submission are authoritative. See
+[the shared production contract](../shared/production-contract.md).
+
+Use current task-relevant published guidance and verified exact endpoint contracts.
+Bundled skills supply production craft; they are not a competing model leaderboard.
+
+## Selection policy
+
+1. Preserve the user’s explicit provider and model choice. Verify the canonical
+   endpoint; do not silently substitute a family or variant.
+2. For a variation, recover the source with `get_job_recipe`. Keep the actual
+   source endpoint and technical inputs ahead of personal model preferences.
+3. Otherwise call `recommend_model` with the task, actual source roles/counts,
+   hard output requirements, budget and preferences. Follow its fit and price
+   evidence. Popularity, a newest release or a model family name is not proof
+   of suitability. `workflow_needed` means plan each step rather than claiming
+   one endpoint satisfies the composition.
+4. `recommend_model` already applies published routing guidance. Read
+   `search_guidance`/`read_guidance` only for multi-step or specialised work,
+   or when the user asks. Keep claim/version identity and retrieval health;
+   unrelated or unreviewed advice does not override a compatible task-specific
+   claim.
+5. When `recommend_model` selects a model, build the input from its
+   `next_call.input_contract` and submit with `run_model` or `submit_job`.
+   Submission validates against the live schema before any charge and returns
+   field errors with expected values. `get_model_schema` is optional detail
+   for parameters the contract omits. Current schema facts resolve
+   conflicting parameter advice. Prompting advice cannot create missing
+   fields, controls, input roles or native resolution.
+6. When the user asks about cost or sets a budget, compare actual priced units: use
+   `next_call.price` when present, otherwise `get_pricing` or `estimate_cost`.
+   Unknown price remains unresolved. Never claim a hard budget is met from an
+   unrelated unit rate.
+7. If there is no supported fit, explain the unresolved requirement. Use live
+   discovery to inspect candidates; do not invent endpoint IDs or hardcode a
+   fallback model ranking.
+
+## Artifact requirements
+
+- Product/character preservation needs actual supported reference inputs and
+  later pixel inspection. An image field does not guarantee fidelity.
+- Readable text needs exact user-provided copy and current task-specific
+  guidance; reserve external typography when generated text cannot pass.
+- Talking heads need script/audio duration and observed sync checks.
+- Multi-shot video needs checked anchors and retained source roles.
+- 3D needs verified format, material, scale and requested rigging capability.
+
+Read [the shared production contract](../shared/production-contract.md) before
+chaining preserved references or claiming final delivery. Execute with MCP
+tools; do not call the genmedia CLI from this bundle.
Full technical diff · 1 changed fields

changed /skill_md_contents

BEFORE
"---\nname: model-routing\ndescription: >\n  Choose production-ready fal.ai endpoint IDs for MCP media workflows. Use\n  with commercial, marketing, ugc, character-design, cinematography,\n  storytelling, fal-recipes, and fal-workflow when the user has not named a\n  specific model or asks which model to use.\n---\n\n# fal.ai Model Routing\n\nUse this skill for curated endpoint choice. It is not a replacement for the\nMCP tools. After choosing an endpoint, execute through the fal.ai MCP tools in\nthis plugin. Do not call the genmedia CLI from this MCP skill bundle.\n\n## Routing Rules\n\n1. If the user did not name a specific endpoint, call `recommend_model` first\n   for a current catalog-ranked recommendation.\n2. Use the curated endpoints in this skill to evaluate the recommendation and\n   pick the artifact-appropriate route. If the recommendation is generic but\n   the task is specialized, use `search_models` to verify a better endpoint.\n3. Use live MCP discovery when the role is not covered, an endpoint is missing\n   or deprecated, or the user asks to compare options.\n4. Inspect the selected endpoint schema before using custom fields.\n5. Check pricing when cost, batch size, video duration, 3D, or high-resolution\n   output matters.\n6. Do not invent endpoint IDs or schema fields.\n\n## Image Generation\n\nGeneric/simple image generation:\n\n- `fal-ai/flux-2/klein/9b`: default when the user asks for a basic image and\n  does not name a model or require premium text rendering, brand fidelity, or\n  final commercial quality.\n\nText-heavy image work:\n\n- `openai/gpt-image-2`: best default for readable text, posters, packaging,\n  diagrams, UI mockups, book covers, exact-copy layouts, and high-end stills.\n  Use high quality when the result is final.\n- `fal-ai/nano-banana-pro`: second choice when text is important but cost or\n  availability matters.\n\nPremium still images:\n\n- `openai/gpt-image-2`: commercial realism, editorial photography, product\n  scenes, character anchors, and high-quality concept art.\n- `fal-ai/nano-banana-pro`: strong styled output and product-friendly images.\n- `fal-ai/nano-banana-2`: cheaper strong alternative.\n\nFast draft images:\n\n- `fal-ai/flux-2/klein/9b`: quick concepts, mood exploration, rough layouts.\n  Do not use as final commercial delivery unless the user asks for speed/cost.\n\n## Image Editing\n\nUse for product edits, background replacement, inpainting, relighting, cleanup,\nobject changes, outfit changes, and reference-driven composition.\n\n1. `fal-ai/flux-2/klein/9b/edit`: default for simple/generic image edits when\n   the user does not name a model.\n2. `fal-ai/nano-banana-pro/edit`\n3. `openai/gpt-image-2/edit`\n4. `fal-ai/bytedance/seedream/v5/lite/edit`\n\nFor product fidelity, prefer reference/edit workflows over text-only generation.\n\n## Video\n\nHighest quality video:\n\n- Text to video: `bytedance/seedance-2.0/text-to-video`\n- Image to video: `bytedance/seedance-2.0/image-to-video`\n- Reference to video: `bytedance/seedance-2.0/reference-to-video`\n\nFast/lower-cost video:\n\n- Text to video: `xai/grok-imagine-video/text-to-video`\n- Image to video: `xai/grok-imagine-video/image-to-video`\n- Video edit: `xai/grok-imagine-video/edit-video`\n\nMulti-shot or control-heavy video:\n\n1. `bytedance/seedance-2.0/text-to-video`\n2. `bytedance/seedance-2.0/image-to-video`\n3. `bytedance/seedance-2.0/reference-to-video`\n4. `fal-ai/kling-video/v3/pro/text-to-video`\n5. `fal-ai/kling-video/v3/pro/image-to-video`\n6. `alibaba/happy-horse/text-to-video`\n7. `alibaba/happy-horse/image-to-video`\n\nUse Kling when multi-prompt, element controls, or stronger prompt structure are\nneeded. Use Happy Horse for brief, direct camera-language prompts.\n\n## UGC And Talking Head\n\n- Portrait plus audio: `veed/fabric-1.0`\n- Portrait plus text: `veed/fabric-1.0/text`\n- Avatar with visual direction: `fal-ai/creatify/aurora`\n- Existing footage with new speech: `fal-ai/sync-lipsync/v2`\n- Product b-roll: `bytedance/seedance-2.0/image-to-video`\n- Fast b-roll draft: `xai/grok-imagine-video/image-to-video`\n\n## Audio\n\nUse live catalog discovery for current TTS, music, sound effect, transcription,\nand cleanup endpoints. Audio model availability changes quickly, so inspect\nschemas and pricing before execution.\n\n## 3D\n\n- Premium image-to-3D or character assets: use Meshy v6 class endpoints when\n  available and inspect rigging/animation fields.\n- Hunyuan/Tripo-style endpoints can be alternatives for faster drafts or\n  different geometry outputs.\n- Always run 3D jobs async and return model URLs plus any texture/material URLs.\n\n## Quality Guardrail\n\nModel choice is only valid if it matches the artifact role:\n\n- Text-heavy deliverables need text-capable image models or external layout.\n- Product deliverables need reference/edit workflows when packaging must hold.\n- Talking-head deliverables need audio/script length checks.\n- Story deliverables need continuity anchors across shots.\n- 3D deliverables need output format, scale, material, and rigging checks.\n"
AFTER
"---\nname: model-routing\ndescription: >\n  Choose production-ready fal.ai endpoint IDs for MCP media workflows. Use\n  with commercial, marketing, ugc, character-design, cinematography,\n  storytelling, fal-recipes, and fal-workflow when the user has not named a\n  specific model or asks which model to use.\n---\n\n# fal.ai Model Routing\n\nBundled guidance is fallback craft. For inputs, recommend_model's\n`input_contract` and live validation at submission are authoritative. See\n[the shared production contract](../shared/production-contract.md).\n\nUse current task-relevant published guidance and verified exact endpoint contracts.\nBundled skills supply production craft; they are not a competing model leaderboard.\n\n## Selection policy\n\n1. Preserve the user’s explicit provider and model choice. Verify the canonical\n   endpoint; do not silently substitute a family or variant.\n2. For a variation, recover the source with `get_job_recipe`. Keep the actual\n   source endpoint and technical inputs ahead of personal model preferences.\n3. Otherwise call `recommend_model` with the task, actual source roles/counts,\n   hard output requirements, budget and preferences. Follow its fit and price\n   evidence. Popularity, a newest release or a model family name is not proof\n   of suitability. `workflow_needed` means plan each step rather than claiming\n   one endpoint satisfies the composition.\n4. `recommend_model` already applies published routing guidance. Read\n   `search_guidance`/`read_guidance` only for multi-step or specialised work,\n   or when the user asks. Keep claim/version identity and retrieval health;\n   unrelated or unreviewed advice does not override a compatible task-specific\n   claim.\n5. When `recommend_model` selects a model, build the input from its\n   `next_call.input_contract` and submit with `run_model` or `submit_job`.\n   Submission validates against the live schema before any charge and returns\n   field errors with expected values. `get_model_schema` is optional detail\n   for parameters the contract omits. Current schema facts resolve\n   conflicting parameter advice. Prompting advice cannot create missing\n   fields, controls, input roles or native resolution.\n6. When the user asks about cost or sets a budget, compare actual priced units: use\n   `next_call.price` when present, otherwise `get_pricing` or `estimate_cost`.\n   Unknown price remains unresolved. Never claim a hard budget is met from an\n   unrelated unit rate.\n7. If there is no supported fit, explain the unresolved requirement. Use live\n   discovery to inspect candidates; do not invent endpoint IDs or hardcode a\n   fallback model ranking.\n\n## Artifact requirements\n\n- Product/character preservation needs actual supported reference inputs and\n  later pixel inspection. An image field does not guarantee fidelity.\n- Readable text needs exact user-provided copy and current task-specific\n  guidance; reserve external typography when generated text cannot pass.\n- Talking heads need script/audio duration and observed sync checks.\n- Multi-shot video needs checked anchors and retained source roles.\n- 3D needs verified format, material, scale and requested rigging capability.\n\nRead [the shared production contract](../shared/production-contract.md) before\nchaining preserved references or claiming final delivery. Execute with MCP\ntools; do not call the genmedia CLI from this bundle.\n"
Full snapshot data
{
  "description": "Choose production-ready fal.ai endpoint IDs for MCP media workflows. Use with commercial, marketing, ugc, character-design, cinematography, storytelling, fal-recipes, and fal-workflow when the user has not named a specific model or asks which model to use.\n",
  "included_files": [],
  "name": "model-routing",
  "skill_md_contents": "---\nname: model-routing\ndescription: >\n  Choose production-ready fal.ai endpoint IDs for MCP media workflows. Use\n  with commercial, marketing, ugc, character-design, cinematography,\n  storytelling, fal-recipes, and fal-workflow when the user has not named a\n  specific model or asks which model to use.\n---\n\n# fal.ai Model Routing\n\nBundled guidance is fallback craft. For inputs, recommend_model's\n`input_contract` and live validation at submission are authoritative. See\n[the shared production contract](../shared/production-contract.md).\n\nUse current task-relevant published guidance and verified exact endpoint contracts.\nBundled skills supply production craft; they are not a competing model leaderboard.\n\n## Selection policy\n\n1. Preserve the user’s explicit provider and model choice. Verify the canonical\n   endpoint; do not silently substitute a family or variant.\n2. For a variation, recover the source with `get_job_recipe`. Keep the actual\n   source endpoint and technical inputs ahead of personal model preferences.\n3. Otherwise call `recommend_model` with the task, actual source roles/counts,\n   hard output requirements, budget and preferences. Follow its fit and price\n   evidence. Popularity, a newest release or a model family name is not proof\n   of suitability. `workflow_needed` means plan each step rather than claiming\n   one endpoint satisfies the composition.\n4. `recommend_model` already applies published routing guidance. Read\n   `search_guidance`/`read_guidance` only for multi-step or specialised work,\n   or when the user asks. Keep claim/version identity and retrieval health;\n   unrelated or unreviewed advice does not override a compatible task-specific\n   claim.\n5. When `recommend_model` selects a model, build the input from its\n   `next_call.input_contract` and submit with `run_model` or `submit_job`.\n   Submission validates against the live schema before any charge and returns\n   field errors with expected values. `get_model_schema` is optional detail\n   for parameters the contract omits. Current schema facts resolve\n   conflicting parameter advice. Prompting advice cannot create missing\n   fields, controls, input roles or native resolution.\n6. When the user asks about cost or sets a budget, compare actual priced units: use\n   `next_call.price` when present, otherwise `get_pricing` or `estimate_cost`.\n   Unknown price remains unresolved. Never claim a hard budget is met from an\n   unrelated unit rate.\n7. If there is no supported fit, explain the unresolved requirement. Use live\n   discovery to inspect candidates; do not invent endpoint IDs or hardcode a\n   fallback model ranking.\n\n## Artifact requirements\n\n- Product/character preservation needs actual supported reference inputs and\n  later pixel inspection. An image field does not guarantee fidelity.\n- Readable text needs exact user-provided copy and current task-specific\n  guidance; reserve external typography when generated text cannot pass.\n- Talking heads need script/audio duration and observed sync checks.\n- Multi-shot video needs checked anchors and retained source roles.\n- 3D needs verified format, material, scale and requested rigging capability.\n\nRead [the shared production contract](../shared/production-contract.md) before\nchaining preserved references or claiming final delivery. Execute with MCP\ntools; do not call the genmedia CLI from this bundle.\n"
}

SHA-256 of public snapshot: f173ba655c20c3021888b1bd75cef818d315c59227fc1d99a453b69d118c1c34