Loading ToolBench
Parallel video rendering with live dashboard, GPU auto-detection, checkpoint system, and stream-copy concat. Includes an MCP server, a Claude Code skill, and a CLI.
Strong tool definitions with clear naming, comprehensive descriptions (avg 180 chars), and well-structured JSON schemas. All 7 tools have input schemas with type constraints and enums where appropriate. Descriptions explain WHAT, WHEN, and prerequisites. However, output schemas are not documented, LLMs cannot predict response structure. Error handling guidance is minimal. Tool composition is sound (each does one thing), but some parameter descriptions lack format/range details. Naming follows verb_noun convention consistently.
Apply color grading presets or custom filters to a video file without re-encoding. Supports built-in presets (cinematic, vintage, cool, warm, high-contrast, desaturate) or custom LUT/filter chains. No video re-encode, only color filter application.
Stream-copy concatenate multiple video files without re-encoding. Joins segments using ffmpeg's concat demuxer for fast, lossless concatenation. All input files must have identical codec, resolution, and frame rate.
Detect the best available hardware video encoder on this machine. Probes NVENC, VideoToolbox, AMF, VA-API, and QSV with a 1-frame validation encode and falls back to CPU (libx264) when no GPU encoder works. Use before rendering or grading to confirm hardware acceleration. Results are cached for 7 days in ~/.ffmpeg-render-pro/gpu-cache.json. Requires ffmpeg on PATH or FFMPEG_RENDER_PRO_FFMPEG.
Get the worker script contract and bundled reference worker source. Returns the starter worker template that can be used as a base for custom render workers, including the renderFrame function signature and the full basic-worker.js reference implementation.
Combine video with audio track without re-encoding the video. Supports audio looping for short audio files, optional loudness normalization, and AAC audio encoding. No video re-encode.
Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination). render_video, color_grade, merge_audio, concat_videos lack documented return types.
render_video parameter descriptions lack format/range details. 'workerScript' and 'outputPath' need path format guidance. 'codecArgs' array lacks element type or example structure.
color_grade 'customFilter' parameter description is vague. Should specify ffmpeg filter syntax, valid operators, and link to ffmpeg filter documentation.
Error handling guidance missing. Tools do not document recovery paths (e.g., 'If ffmpeg not found, set FFMPEG_RENDER_PRO_FFMPEG env var'). No categorization of retryable vs fatal errors.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2025-06-18+ | v2 |
Parallel render with progress notifications and cancellation support. Renders a video using a worker script, distributing frames across parallel workers, with live dashboard, checkpoint recovery, and stream-copy concatenation. Requires ffmpeg on PATH or FFMPEG_RENDER_PRO_FFMPEG.
Get render system capabilities: CPU cores, total and free RAM, the recommended parallel worker count for a target resolution, detected GPU encoder, segment and final codecs, and ffmpeg version. Use before render_video to choose a worker count or check whether this machine can handle a resolution. Requires ffmpeg on PATH or FFMPEG_RENDER_PRO_FFMPEG.
render_video lacks confirmation/dry-run pattern for destructive operation. Parallel rendering with checkpoint recovery is complex; agents should preview plan before execution.