runware-models
Skill8 starsChanged 5 months agoMITView on GitHub
Pick the right Runware model for a task and keep that choice current. Use when an agent needs to decide "which model for X", discover what's available, or check a model's capabilities, status, or price. Foundation skill - outcome skills delegate model choice here so nothing rots when the catalog changes.
In this repository32 other files it gives its agents.
- add-audio-to-videoSkill
- animate-imageSkill
- bring-your-own-modelSkill
- character-consistencySkill
- cinematic-videoSkill
- composite-sceneSkill
- controlled-generationSkill
- dialogue-audioSkill
- edit-imageSkill
- edit-videoSkill
- game-assets-2dSkill
- image-to-3d-assetSkill
- llm-agentSkill
- logos-and-vectorsSkill
- multi-shot-videoSkill
- musicSkill
- photoreal-stillsSkill
- product-ad-videoSkill
- product-photographySkill
- reframe-videoSkill
- replace-in-videoSkill
- restore-and-upscaleSkill
- runware-promptingSkill
- runware-runSkill
- talking-avatarSkill
- text-in-imageSkill
- train-style-modelSkill
- ugc-adSkill
- video-upscaleSkill
- vision-understandingSkill
- voice-cloningSkill
- voiceoverSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 30
Cost-aware: Check the card's pricing (or dry-run via runware-run) before a batch.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Choosing a Runware model
Runware's catalog is large and changes weekly. The rule: never hardcode a single model as the only answer. Start from a sensible default for the capability, then confirm it's live and current with a real lookup before running.
How to discover
- List / search the catalog (SDK, MCP
list_models/model_search, or CLI) filtered by capability and modality. Each model card carriesname,air,status,capabilities, pricing, and a headline. - Capabilities are tagged as
io:<from>-to-<to>andop:<operation>:io:text-to-image,io:image-to-image,io:text-to-video,io:image-to-video,io:video-to-video,io:text-to-audio,io:audio-to-video(lipsync),io:image-to-3d,io:text-to-3d,io:text-to-text,io:image-to-text(vision).op:edit,op:remove-background,op:upscale,op:extend,op:vectorize,op:caption,op:mask,op:train, and the ControlNet preprocessors (op:depth-estimation,op:pose-estimation, …).
- Match the customer's outcome to a capability, then to a model.
Status matters
status tells you whether a model is usable: prefer live. Treat deprecated as replaced (check replacedBy), coming-soon / draft as not yet callable, unlisted as reachable only if you already have its ID.
Routing principles
- Default to a flagship for quality, offer a fast/cheap tier for drafts. Most families have both (e.g. a
[max]/protier and a fastklein/flash/turbotier). - Route by sub-task, not by brand. A single outcome often spans models: e.g. text-heavy images favor a model strong at typography; a premium still vs a fast draft are different picks.
- Cost-aware. Check the card's pricing (or dry-run via
runware-run) before a batch. - Hosted models carry a "Runware Optimized" marker - good defaults for reliability/price.
Quality bar
- The chosen model is
liveand actually supports the required capability (verified against its card, not assumed). - A cheaper/faster alternative was considered for drafts or high-volume runs.
Related skills
runware-run (call the model you picked), runware-prompting (prompt it well), and every outcome skill (each names its default routing, then defers here to confirm).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

