Z Image Turbo

Tools

• ImageStudio_apply_model_defaults Update sliders and the negative-prompt controls when the model changes. Sampler defaults + capabilities come from the model's registry entry, so this stays correct for any model the registry knows about.
• ImageStudio_assistant_chat Gradio-facing wrapper around ``vlm_chat``. Yields ``(answer, progress)`` so the UI streams the text live and a downstream consumer reading the progress index sees this node advance over SSE. The final ``complete`` frame carries the clean answer at index 0. ``system_prompt`` (optional) steers the assistant. If a user passes a file as an input, use the upload_file_to_gradio tool, if present, to upload the file to the gradio app and create a Gradio File Input. Then use the returned path as the input to the tool
• ImageStudio_assistant_chat_1 Gradio-facing wrapper around ``vlm_chat``. Yields ``(answer, progress)`` so the UI streams the text live and a downstream consumer reading the progress index sees this node advance over SSE. The final ``complete`` frame carries the clean answer at index 0. ``system_prompt`` (optional) steers the assistant. If a user passes a file as an input, use the upload_file_to_gradio tool, if present, to upload the file to the gradio app and create a Gradio File Input. Then use the returned path as the input to the tool
• ImageStudio_generate_and_upload Generate, then upload the result to R2 outside the GPU window. Generator yielding ``(image, seed, r2_status, progress)``. Streams per-step sampling progress (image still None) and finishes with the real image, seed and R2 status once the upload completes. The image is always the original HF-generated asset; ``r2_status`` reports the uploaded filekey, bucket and a presigned ``r2_url`` on success (or the error on failure) plus the VLM ``moderation`` result. The asset is encoded as **WebP** before upload (and the returned file is WebP too) so downstream consumers only ever see WebP. The caller's unique id (``uid`` cookie) is recorded in the object's metadata.
• ImageStudio_generate_and_upload_1 Generate, then upload the result to R2 outside the GPU window. Generator yielding ``(image, seed, r2_status, progress)``. Streams per-step sampling progress (image still None) and finishes with the real image, seed and R2 status once the upload completes. The image is always the original HF-generated asset; ``r2_status`` reports the uploaded filekey, bucket and a presigned ``r2_url`` on success (or the error on failure) plus the VLM ``moderation`` result. The asset is encoded as **WebP** before upload (and the returned file is WebP too) so downstream consumers only ever see WebP. The caller's unique id (``uid`` cookie) is recorded in the object's metadata.
• ImageStudio_image_to_video_assets User-uploaded image -> motion prompt + image-context (moderation) check. Generator yielding ``(video_prompt, status, progress)`` per frame so the motion prompt streams and a downstream consumer reading the progress index sees live progress over SSE. ``status`` carries the moderation check once available; the final ``complete`` frame holds the clean prompt at index 0. Bound with ``api_name="image_to_video_assets"`` (hidden trigger, no UI). If a user passes a file as an input, use the upload_file_to_gradio tool, if present, to upload the file to the gradio app and create a Gradio File Input. Then use the returned path as the input to the tool
• ImageStudio_moderate_image Standalone moderation endpoint (no UI) for evaluating/tuning the guard. Runs the image-context safety check on a user-supplied image and returns the parsed moderation dict (same shape as the inline checks). Lets us A/B the actual guard model (Qwen 3.5) against the assistant VLM, and try alternate ``MODERATION_PROMPT`` wordings, against fixed images without redeploying for each variant. Bound with api_name="moderate_image". If a user passes a file as an input, use the upload_file_to_gradio tool, if present, to upload the file to the gradio app and create a Gradio File Input. Then use the returned path as the input to the tool
• ImageStudio_prompt_to_video_assets Text -> first-frame image (uploaded to R2) + video generation prompt. Streams progress and returns a dict with ``video_prompt``, ``first_frame_url`` (a presigned GET to the **WebP** first frame, usable directly), ``r2_filekey``/``r2_bucket`` (for callers that resolve their own public URL), ``first_frame_prompt`` (the intermediate image prompt), ``moderation`` (the VLM image-context check on the rendered frame) and ``seed``. Has no UI (registered via ``gr.api``).

No tool matches.

As the server's own listing gives them. We never call a tool.