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/asr-api.md
View on GitHub# Doubao recording recognition business fields
Submit the native business body through the model proxy with `api=doubao.asr`. Follow `mcp.md` for MCP argument mapping. Do not wrap the body in a native HTTP envelope or add authentication headers, resource IDs, request IDs or provider query instructions as body fields. The proxy controls service access; `model_name=bigmodel` alone does not select model generation 1.0 versus 2.0.
## Default request
```json
{
"audio": {
"url": "https://example.com/separated-voice.wav",
"format": "wav"
},
"request": {
"model_name": "bigmodel",
"enable_itn": true,
"enable_punc": true,
"enable_ddc": false,
"show_utterances": true
}
}
```
Replace the example URL with the actual separation output. No top-level `model`, `prompt`, `text`, `speaker` or `user.uid` is required by this request contract.
## Audio fields
| Field | Rule |
| --- | --- |
| `audio.url` | Required accessible HTTP(S) URL to the separated voice file; never Base64 or a local path |
| `audio.format` | Required actual format: `wav`, `mp3`, `ogg`, `spx`, `amr`, `aac`, or `m4a` |
| `audio.language` | Optional supported language code; leave unset for the documented Chinese/English/dialect recognition mode |
| `audio.codec` | Optional `raw` or `opus`; only set from known encoding |
| `audio.rate` | Optional sample rate; native default 16000, not evidence about a particular file |
| `audio.bits` | Optional bit depth; native default 16 |
| `audio.channel` | Optional 1 or 2; native default 1 |
Use WAV separation output to stay within the formats explicitly supported in both the endpoint overview and its parameter table. The overview names WAV/MP3/OGG, while the field table lists additional formats; do not broaden the default based on that discrepancy. The recognition limit is 512 MB and five hours for the submitted file.
Language codes listed by the submission documentation include `zh-CN`, `en-US`, `ja-JP`, `id-ID`, `es-MX`, `pt-BR`, `de-DE`, `fr-FR`, `ko-KR`, `fil-PH`, `ms-MY`, `th-TH`, `ar-SA`, `it-IT`, `bn-BD`, `el-GR`, `nl-NL`, `ru-RU`, `tr-TR`, `vi-VN`, `pl-PL`, `ne-NP`, `uk-UA`, and `yue-CN`. The Romanian code is inconsistent between the explicit-language and automatic-language tables; do not guess a forced code when it has not been confirmed. Unset language covers Chinese, English, Shanghainese, Minnan, Sichuan, Shaanxi and Cantonese; it does not imply unrestricted automatic language detection.
## Recognition options
| `request` field | Usage |
| --- | --- |
| `model_name` | Required `bigmodel` |
| `show_utterances` | Set true to obtain available utterance/word/timing details; provider default false |
| `enable_punc` | Punctuation, default true |
| `enable_itn` | Normalize spoken numbers, money and dates, default true; disable for literal number wording |
| `enable_ddc` | Semantic smoothing, default false; leave false to avoid intentional removal/correction of fillers and repetitions |
| `enable_speaker_info` | Optional diarization, default false; requires show_utterances and language unset or zh-CN |
| `ssd_version` | Optional diarization model: `200` for non-meeting scenarios (suggested at most five speakers), `300` for long multi-speaker meetings; requires speaker and utterance options |
| `ssd_mode` | Only version 200: 0 ordinary short audio, 1 clustering for longer non-meeting audio |
| `enable_channel_split` | Optional stereo-channel recognition, default false; not interchangeable with speaker diarization; use only if the separated file retains meaningful channels |
| `enable_auto_lang` | Optional broader automatic language recognition, default false; incompatible with corpus/context configuration |
| `corpus.context` | Optional JSON-serialized string containing supplied hotwords or context; not a nested object |
Add only requested or task-relevant options. Do not automatically infer age, gender, emotion, or turn on content filters. For user-supplied hotwords, an example is `"corpus":{"context":"{\"hotwords\":[{\"word\":\"MediaKit\"}]}"}` under `request`. Total hotwords are limited to 5000, and context to 500 tokens. Hotwords guide recognition; they do not justify inserting inaudible words. Do not enable auto-language alongside corpus.
## Recognition result
The native successful query example contains `audio_info.duration` and a `result` object with `text` and optional `utterances`. Each utterance can include `text`, `start_time`, `end_time`, `words`, and `additions`; speaker/channel labels, when enabled, appear in additions. Word entries may contain text, timings and confidence. Preserve actual values instead of assuming every field is present.
Durations and start/end times in this response are milliseconds. Missing word timings are not an error if sentence-level output suffices. Missing utterances mean timed subtitles cannot be promised. Speaker labels are task-local identifiers.
Native success is represented by status `20000000` when surfaced, but the proxy's response may expose status differently. Inspect the actual service result and transcript rather than requiring undocumented wrapper fields or treating HTTP/outer success as recognition success. Do not parse a submission acknowledgment as a query result.
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 |