yikart/AiToEarn / skills/codex-plugin-aitoearn/skills/aitoearn/references/byted-mediakit-asr
byted-mediakit-asr
VerifiedExtract speech subtitles from videos via MediaKit MCP. Only for full transcripts, use voice separation followed by Doubao ASR through call-maas-proxy on audio or video. Use for subtitle extraction, SRT/VTT, recording-to-text and timestamped transcripts. Not for speech synthesis, visual OCR or general sound description.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.0 KB |
| evals/evals.json | config | 3.6 KB |
| references/asr-api.md | markdown | 5.1 KB |
| references/mcp.md | markdown | 3.5 KB |
references/mcp.md
View on GitHub# MCP execution
## Discover and map arguments
Discover the callable model-proxy capability by exposed name, description and public schema, commonly `call_maas_proxy` (prefixes and spelling may vary). Confirm it supports `doubao.asr`; do not substitute an unrelated model tool. Use the actual discovered name. The host resolves the submitted task and returns its completed result. If this proxy capability is unavailable, report the gap.
Build the business request using this skill's bundled API reference. Select `doubao.asr` through the proxy's exposed API selector. Pass the business object through its documented parameter field (commonly `maas_params`): use a JSON object if the schema accepts an object, or serialize once if it requires a string. Do not copy CLI flags, shell quoting, CLI-only controls or provider authentication headers into MCP arguments. Omit optional fields that are not needed and wrapper-fixed parameters.
## Inputs
Media references inside the model business request must resolve to accessible HTTP(S) URLs. Do not send Base64, data URLs, local paths or `mediakit://` handles as model parameters. Use MCP media references only in fields whose public schema and resolver explicitly support them and resolve them before model submission; do not assume arbitrary nested request fields are resolved. Obtain URLs for local inputs through an available authorized upload route. Use the environment's existing media reference route. If no route can provide a supported reference, report that missing capability.
Preserve signed URLs exactly and never expose credentials. When transferring results between stages, use a reference accepted by the receiving tool.
## Task lifecycle and delivery
Save the selected interface, business request and actual response in the skill's task workspace. Submit the model-proxy request as an asynchronous task and retain the actual returned `task_id`. The host resolves that same task and returns its completed result. A submission without a usable task ID is not a completed result: inspect the response and report the submission problem without inventing an ID or blindly resubmitting. A pending status, timeout, empty output or business error does not establish that MCP is unavailable.
Inspect the actual business result, including nested errors, rather than equating outer success with usable media or recognized text. Apply the result interpretation, saving, quality review and delivery rules from this skill. Do not regenerate or re-recognize to fix a download or display failure.
## Recognition stages
Choose the deliverable before tools. For subtitle-only requests, prefer the discovered subtitle recognition capability, commonly `asr_subtitles`, and follow its actual input schema. Use audio only if that MCP capability explicitly supports audio. Use `asr_subtitles` for subtitle-only requests. Do not substitute voice separation plus Doubao ASR just because the requested input is unsupported.
For full transcripts, prefer the discovered voice-separation capability, commonly `separate_voice`, before the `doubao.asr` proxy. Supply one source in the field for its actual media type, selecting WAV output and the general Audio scene when those options are exposed. Do not invent hidden options. Retain the successful vocal output (commonly `voice_audio_url`); never use the background output or original mixture as a fallback. Await separation before recognition. Reuse adequate separation for the same source.
Use MCP/REST for subtitle recognition, separation and the model proxy. Preserve distinct task IDs and source provenance across stages.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |