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.
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.
Up to 3 plugins free · no card required.
All fields and original values.
Skill instructions
Sep 30, 2026 · 22:46 UTC
This is a curated routing layer for MCP clients. It should guide endpoint selection, Use `model-routing` first for common production defaults. Use this skill when execution and use catalog search when the recommended list is too gener...
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). This is a modality map for MC...
Compare saved observations
Download comparison JSONSKILL.md text changes
− removed · + added
--- before +++ after @@ -8,11 +8,15 @@ # fal.ai Models Catalog -This is a curated routing layer for MCP clients. It should guide endpoint selection, +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). + +This is a modality map for MCP clients. It should guide endpoint selection, then execution should happen through the fal.ai MCP tools exposed by this plugin. Do not shell out to genmedia CLI. -Use `model-routing` first for common production defaults. Use this skill when +Use `model-routing` first for the current selection policy. Use this skill when the question is broader: comparing modalities, finding a category, choosing a family, or explaining tradeoffs. @@ -40,35 +44,29 @@ 1. Identify the artifact role: final commercial, draft, utility transform, analysis, training, or intermediate step. 2. If the user did not name a specific endpoint, call `recommend_model` before - execution and use catalog search when the recommended list is too generic. + execution. When its `decision_status` is `selected`, submit its `next_call`. + Use catalog search only for other statuses. 3. Choose modality and endpoint family. -4. Inspect schema before assuming fields such as `image_url`, `image_urls`, - `reference_image_url`, `duration`, `aspect_ratio`, `seed`, `quality`, - `audio_url`, or `enable_rigging`. -5. Check pricing when the job is long, high resolution, batched, video, audio, - 3D, or likely to be repeated. +4. Use only fields listed in `input_contract`. Never assume fields such as + `image_url`, `image_urls`, `reference_image_url`, `duration`, + `aspect_ratio`, `seed`, `quality`, `audio_url`, or `enable_rigging`. For a + model you picked yourself, read its schema with `get_model_schema`. +5. Look up price only when the user asks about cost or sets a budget. 6. For uncertain categories, use MCP catalog search and docs search, then choose from verified endpoints. -## Production Defaults +## Current production routes -- Text-heavy stills: `openai/gpt-image-2`. -- Product stills and campaign heroes: `openai/gpt-image-2`, - `fal-ai/nano-banana-pro`, `fal-ai/nano-banana-2`. -- Product/reference edits: `fal-ai/nano-banana-pro/edit`, - `openai/gpt-image-2/edit`. -- Final-quality video: Seedance 2.0 text/image/reference video endpoints. -- Fast video drafts: Grok Imagine Video endpoints. -- Talking head: `veed/fabric-1.0`, `veed/fabric-1.0/text`, - `fal-ai/creatify/aurora`, `fal-ai/sync-lipsync/v2`. -- Background removal: use catalog search for current Bria/background endpoints - and inspect schema. -- 3D: prefer Meshy v6 class endpoints for rigging/animation when available. +`recommend_model` applies live task-specific published guidance, exact +endpoint contracts and prices. No bundled endpoint list is a quality ranking or a price promise. Inspect +actual source roles and requested output; a reference-to-video route can have +different constraints from its text-to-video sibling. Preserve version and +retrieval information, and report absent guidance honestly. ## Avoid - Choosing a model only because it is popular when the artifact role is clear. - Using text-to-image when the user supplied a reference that must be preserved. -- Using cheap draft endpoints for final brand/product assets. +- Calling an uninspected draft a finished brand/product asset. - Generating readable legal, medical, financial, or claim text unless the user supplies exact wording and the chosen model supports text well.
Full technical diff · 1 changed fields
changed /skill_md_contents
"---\nname: fal-models-catalog\ndescription: >\n Navigate fal.ai model families by media modality and production role. Use\n when the user asks which model, endpoint, or model family is appropriate for\n image, video, audio, 3D, editing, training, restoration, try-on, or analysis.\n---\n\n# fal.ai Models Catalog\n\nThis is a curated routing layer for MCP clients. It should guide endpoint selection,\nthen execution should happen through the fal.ai MCP tools exposed by this\nplugin. Do not shell out to genmedia CLI.\n\nUse `model-routing` first for common production defaults. Use this skill when\nthe question is broader: comparing modalities, finding a category, choosing a\nfamily, or explaining tradeoffs.\n\n## Modality Map\n\n- Text to image: campaign visuals, product stills, character concepts,\n editorial photography, posters, UI mockups, image typography.\n- Image to image: edits, inpainting, reference preservation, style transfer,\n product placement, background replacement, upscaling, restoration.\n- Text to video: cinematic clips, product reveals, narrative shots, social\n concepts, motion drafts.\n- Image to video: animate an approved still, product hero motion, b-roll,\n talking-head source frames, first-frame continuity.\n- Reference to video: stronger continuity from characters, products, or style\n references where supported.\n- Text/audio to talking head: spokesperson, UGC creator, avatar, lip sync.\n- Text to audio: narration, TTS, music, sound effects.\n- Audio to text: transcription, subtitles, diarization, audio cleanup.\n- Image/text to 3D: objects, characters, game assets, GLB/OBJ/PLY outputs.\n- Image to text / vision: OCR, captioning, segmentation, detection, analysis.\n- Training: LoRA and fine-tune style workflows, only when a dataset exists.\n\n## Selection Pattern\n\n1. Identify the artifact role: final commercial, draft, utility transform,\n analysis, training, or intermediate step.\n2. If the user did not name a specific endpoint, call `recommend_model` before\n execution and use catalog search when the recommended list is too generic.\n3. Choose modality and endpoint family.\n4. Inspect schema before assuming fields such as `image_url`, `image_urls`,\n `reference_image_url`, `duration`, `aspect_ratio`, `seed`, `quality`,\n `audio_url`, or `enable_rigging`.\n5. Check pricing when the job is long, high resolution, batched, video, audio,\n 3D, or likely to be repeated.\n6. For uncertain categories, use MCP catalog search and docs search, then\n choose from verified endpoints.\n\n## Production Defaults\n\n- Text-heavy stills: `openai/gpt-image-2`.\n- Product stills and campaign heroes: `openai/gpt-image-2`,\n `fal-ai/nano-banana-pro`, `fal-ai/nano-banana-2`.\n- Product/reference edits: `fal-ai/nano-banana-pro/edit`,\n `openai/gpt-image-2/edit`.\n- Final-quality video: Seedance 2.0 text/image/reference video endpoints.\n- Fast video drafts: Grok Imagine Video endpoints.\n- Talking head: `veed/fabric-1.0`, `veed/fabric-1.0/text`,\n `fal-ai/creatify/aurora`, `fal-ai/sync-lipsync/v2`.\n- Background removal: use catalog search for current Bria/background endpoints\n and inspect schema.\n- 3D: prefer Meshy v6 class endpoints for rigging/animation when available.\n\n## Avoid\n\n- Choosing a model only because it is popular when the artifact role is clear.\n- Using text-to-image when the user supplied a reference that must be preserved.\n- Using cheap draft endpoints for final brand/product assets.\n- Generating readable legal, medical, financial, or claim text unless the user\n supplies exact wording and the chosen model supports text well.\n"
"---\nname: fal-models-catalog\ndescription: >\n Navigate fal.ai model families by media modality and production role. Use\n when the user asks which model, endpoint, or model family is appropriate for\n image, video, audio, 3D, editing, training, restoration, try-on, or analysis.\n---\n\n# fal.ai Models Catalog\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\nThis is a modality map for MCP clients. It should guide endpoint selection,\nthen execution should happen through the fal.ai MCP tools exposed by this\nplugin. Do not shell out to genmedia CLI.\n\nUse `model-routing` first for the current selection policy. Use this skill when\nthe question is broader: comparing modalities, finding a category, choosing a\nfamily, or explaining tradeoffs.\n\n## Modality Map\n\n- Text to image: campaign visuals, product stills, character concepts,\n editorial photography, posters, UI mockups, image typography.\n- Image to image: edits, inpainting, reference preservation, style transfer,\n product placement, background replacement, upscaling, restoration.\n- Text to video: cinematic clips, product reveals, narrative shots, social\n concepts, motion drafts.\n- Image to video: animate an approved still, product hero motion, b-roll,\n talking-head source frames, first-frame continuity.\n- Reference to video: stronger continuity from characters, products, or style\n references where supported.\n- Text/audio to talking head: spokesperson, UGC creator, avatar, lip sync.\n- Text to audio: narration, TTS, music, sound effects.\n- Audio to text: transcription, subtitles, diarization, audio cleanup.\n- Image/text to 3D: objects, characters, game assets, GLB/OBJ/PLY outputs.\n- Image to text / vision: OCR, captioning, segmentation, detection, analysis.\n- Training: LoRA and fine-tune style workflows, only when a dataset exists.\n\n## Selection Pattern\n\n1. Identify the artifact role: final commercial, draft, utility transform,\n analysis, training, or intermediate step.\n2. If the user did not name a specific endpoint, call `recommend_model` before\n execution. When its `decision_status` is `selected`, submit its `next_call`.\n Use catalog search only for other statuses.\n3. Choose modality and endpoint family.\n4. Use only fields listed in `input_contract`. Never assume fields such as\n `image_url`, `image_urls`, `reference_image_url`, `duration`,\n `aspect_ratio`, `seed`, `quality`, `audio_url`, or `enable_rigging`. For a\n model you picked yourself, read its schema with `get_model_schema`.\n5. Look up price only when the user asks about cost or sets a budget.\n6. For uncertain categories, use MCP catalog search and docs search, then\n choose from verified endpoints.\n\n## Current production routes\n\n`recommend_model` applies live task-specific published guidance, exact\nendpoint contracts and prices. No bundled endpoint list is a quality ranking or a price promise. Inspect\nactual source roles and requested output; a reference-to-video route can have\ndifferent constraints from its text-to-video sibling. Preserve version and\nretrieval information, and report absent guidance honestly.\n\n## Avoid\n\n- Choosing a model only because it is popular when the artifact role is clear.\n- Using text-to-image when the user supplied a reference that must be preserved.\n- Calling an uninspected draft a finished brand/product asset.\n- Generating readable legal, medical, financial, or claim text unless the user\n supplies exact wording and the chosen model supports text well.\n"
Full snapshot data
{
"description": "Navigate fal.ai model families by media modality and production role. Use when the user asks which model, endpoint, or model family is appropriate for image, video, audio, 3D, editing, training, restoration, try-on, or analysis.\n",
"included_files": [],
"name": "fal-models-catalog",
"skill_md_contents": "---\nname: fal-models-catalog\ndescription: >\n Navigate fal.ai model families by media modality and production role. Use\n when the user asks which model, endpoint, or model family is appropriate for\n image, video, audio, 3D, editing, training, restoration, try-on, or analysis.\n---\n\n# fal.ai Models Catalog\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\nThis is a modality map for MCP clients. It should guide endpoint selection,\nthen execution should happen through the fal.ai MCP tools exposed by this\nplugin. Do not shell out to genmedia CLI.\n\nUse `model-routing` first for the current selection policy. Use this skill when\nthe question is broader: comparing modalities, finding a category, choosing a\nfamily, or explaining tradeoffs.\n\n## Modality Map\n\n- Text to image: campaign visuals, product stills, character concepts,\n editorial photography, posters, UI mockups, image typography.\n- Image to image: edits, inpainting, reference preservation, style transfer,\n product placement, background replacement, upscaling, restoration.\n- Text to video: cinematic clips, product reveals, narrative shots, social\n concepts, motion drafts.\n- Image to video: animate an approved still, product hero motion, b-roll,\n talking-head source frames, first-frame continuity.\n- Reference to video: stronger continuity from characters, products, or style\n references where supported.\n- Text/audio to talking head: spokesperson, UGC creator, avatar, lip sync.\n- Text to audio: narration, TTS, music, sound effects.\n- Audio to text: transcription, subtitles, diarization, audio cleanup.\n- Image/text to 3D: objects, characters, game assets, GLB/OBJ/PLY outputs.\n- Image to text / vision: OCR, captioning, segmentation, detection, analysis.\n- Training: LoRA and fine-tune style workflows, only when a dataset exists.\n\n## Selection Pattern\n\n1. Identify the artifact role: final commercial, draft, utility transform,\n analysis, training, or intermediate step.\n2. If the user did not name a specific endpoint, call `recommend_model` before\n execution. When its `decision_status` is `selected`, submit its `next_call`.\n Use catalog search only for other statuses.\n3. Choose modality and endpoint family.\n4. Use only fields listed in `input_contract`. Never assume fields such as\n `image_url`, `image_urls`, `reference_image_url`, `duration`,\n `aspect_ratio`, `seed`, `quality`, `audio_url`, or `enable_rigging`. For a\n model you picked yourself, read its schema with `get_model_schema`.\n5. Look up price only when the user asks about cost or sets a budget.\n6. For uncertain categories, use MCP catalog search and docs search, then\n choose from verified endpoints.\n\n## Current production routes\n\n`recommend_model` applies live task-specific published guidance, exact\nendpoint contracts and prices. No bundled endpoint list is a quality ranking or a price promise. Inspect\nactual source roles and requested output; a reference-to-video route can have\ndifferent constraints from its text-to-video sibling. Preserve version and\nretrieval information, and report absent guidance honestly.\n\n## Avoid\n\n- Choosing a model only because it is popular when the artifact role is clear.\n- Using text-to-image when the user supplied a reference that must be preserved.\n- Calling an uninspected draft a finished brand/product asset.\n- Generating readable legal, medical, financial, or claim text unless the user\n supplies exact wording and the chosen model supports text well.\n"
}SHA-256 of public snapshot: dbcfb2010000433ed14a4c9ba5f6a2cb6213c81635439a867d90e9adb8a18cee