HF Assets
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| HF_TOKEN | No | Free Hugging Face token (Read). A token raises the ZeroGPU quota. It is also read from a .env file next to the server. Never commit it. | |
| FFMPEG_PATH | No | Path to ffmpeg, for reading and writing OGG, MP3 and FLAC. Found on PATH otherwise. | |
| HF_ASSETS_ROOT | No | The directory relative paths are resolved against. Defaults to the working directory. | |
| HF_ASSETS_BLENDER | No | Path to the Blender executable, for render_mesh. Found automatically in the usual install locations otherwise. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| resources | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_modelsA | Every model this server can reach, with its Space, licence and rough cost. Read this before generating: it says which models may ship and which are prototype-only. |
| generate_imageA | Generate an image from a text prompt through a free Hugging Face Space. Writes the image and a .json sidecar with the prompt, the parameters, the seed and the raw Space response. Use klein with reference_images to edit an existing picture, qwen when the image has to carry legible text. |
| generate_meshA | Turn an image into a raw 3D mesh. trellis gives a textured GLB at ~30k faces; hunyuan gives an untextured shape at ~450k faces. Writes the GLB and a .json sidecar with the parameters, the inputs and the raw Space response. |
| generate_audioA | Generate sound effects, music or speech through a free Hugging Face Space. Writes the audio and a .json sidecar with the prompt, the parameters, the seed and the raw Space response. mmaudio is CC-BY-NC-4.0: prototype only, it must never reach a build. |
| generate_audio_batchA | Generate every asset in a manifest, two takes each, with its sidecar. The manifest lists the prompt, the model, the parameters, the licence and the target loudness of each asset. The tool refuses a manifest that names a model with a non-commercial licence, so nothing that must not ship gets generated in the first place. |
| postprocess_audioA | Trim the silence, crossfade the ends into a seamless loop, normalise the level. This is the step that turns a chosen take into a finished asset. With loop on, the file is shortened by the crossfade so that it tiles without a click at the seam. |
| synth_soundA | Synthesise audio without the network and without spending any quota. engine writes a loop whose length is a whole number of firing cycles, so it still loops cleanly when the game pitches it up or down. sfx writes vehicle Foley: crash, tyre skid and boost. music writes a breakbeat in four intensity tiers plus a menu theme. |
| model_sheetA | Assemble a 4-in-1 model sheet prompt: front, rear, side and 3/4 in one image. One prompt, one board, four angles of the same object — the fastest way to get a consistent set of views, and the best input for the image-to-3D models. The object preset suits any prop, building, character or machine; the vehicle preset keeps a car or a bike in factory trim. |
| render_meshA | Render a mesh from six studio angles: both ends, both sides and two 3/4 views. The mesh is centred, put on the floor and lit with a key, a fill and a rim light. Use it to judge the shape before building anything on top of it. Needs Blender; set HF_ASSETS_BLENDER if it is not on PATH. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| prompt-templates | The canonical prompt templates for concepts and model sheets. |
| model-catalogue | Every model with its Space, licence and cost. |
| audio-manifest | An example audio manifest, showing the format the batch tool expects. |
TDQS
Scored across 9 tools
Most tools have distinct purposes, but the audio tools (generate_audio, generate_audio_batch, synth_sound, postprocess_audio) share a broad domain and require careful reading to pick correctly. model_sheet could also be mistaken for an image generator, though its description clarifies it assembles a prompt.
Eight of nine tools follow a consistent snake_case verb_noun pattern (list_models, generate_image, generate_mesh, generate_audio, generate_audio_batch, postprocess_audio, synth_sound, render_mesh). model_sheet is a noun phrase, which is a minor deviation from the otherwise strong convention.
Nine tools are well-scoped for an asset-generation pipeline covering images, meshes, audio, and supporting steps. Each tool has a distinct job, and the set is neither thin nor bloated.
The core lifecycle for generating, post-processing, and rendering image, mesh, and audio assets is well covered, including model and licence awareness. Minor gaps exist around inspecting or managing generated artifacts and image post-processing, but they do not block the main workflows.