turbot2av-extension
Plan TurboT2AV text-to-audio-video extension setup, checkpoints, Pixi/LTX environment separation, inference commands, and TurboDiffusion-style acceleration choices.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- turbot2av-extension
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Plan TurboT2AV text-to-audio-video extension setup, checkpoints, Pixi/LTX environment separation, inference commands, and TurboDiffusion-style acceleration choices.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/acceleration-notes.md8.1 KB
- 参考文档references/troubleshooting.md11.3 KB
- 参考文档references/turbot2av.md8.1 KB
03 SKILL.md 正文
TurboT2AV Extension
Use this sub-skill when the task is about the TurboT2AV text-to-audio-video extension rather than the core Wan T2V/I2V TurboDiffusion workflows.
Read these bundled references before giving commands:
- TurboT2AV setup and inference for environment separation, required weights, prompt files, student/teacher inference planning, and CLI options.
- Acceleration notes for SageSLA, TileLang W8A8, FastNorm,
topk=0.3,trim_text_context, and how TurboT2AV acceleration differs from core TurboDiffusion. - Troubleshooting for Pixi/LTX isolation, Gemma gated access, missing checkpoints, CUDA/TileLang/SageAttention/SpargeAttn failures, and interpretation pitfalls.
Use the bundled command renderer instead of hand-assembling long inference commands:
- scripts/build_turbot2av_command.py renders a student or teacher
ltx_distillation.tools.run_av_inference_evalcommand from user-supplied paths and options. It validates that required paths are supplied, prints only commands, and never downloads weights or runs inference.
Route elsewhere:
- Core Wan text-to-video or image-to-video inference belongs to the sibling video-inference guidance, not this sub-skill.
- TurboDiffusion CUDA operator build and Wan-specific
quant_lineartroubleshooting belongs to the acceleration-backends guidance; this sub-skill only explains how those ideas are adapted for LTX/TurboT2AV. - Full LTX-2 internals, training, and Pixi installation script maintenance are reference-only; do not duplicate or rewrite the LTX-2 project here.
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.