aidso-human-video-generator
当用户需要绑定 AIDSO、提取参考视频文案,或用用户提供的文案生成并查询数字人视频时使用。
Aidso Human Video Generator
Overview
Use this skill to bind an AIDSO account, extract benchmark-video text, and generate digital-human videos through AIDSO Skills API. Do not import, paste, bundle, or display user secrets. On first use, start the AIDSO authorization-link flow and save only the returned local binding file.
Run the commands below from the directory containing this SKILL.md. When the shell starts elsewhere, use an absolute path to each script in this skill's scripts/ directory.
Keep installation and package-delivery output platform-agnostic. Do not report host-specific installation or local binding state in that output. Account binding is handled only when the skill is first used or when the user explicitly asks to check authorization.
Workflow
- Ensure runtime authorization. If no local binding exists, follow the two-phase authorization flow below and stop until the user approves it.
- Collect the digital-human description without requiring body-framing details.
- Ask the user to choose the generation method: imitation or original with their own video script.
- Prepare a script source and stop for user confirmation.
- Collect video configuration: subtitle
on|off. - Ask whether the user wants a preview or direct full generation.
- For full generation, get the estimated points and explicit user confirmation.
- Submit the video generation task.
- Return the
taskIdto the user. - Poll the result only when requested or when the user wants the final links.
Generation Methods
Use one of these paths.
Imitation
- Ask for the benchmark video URL.
- Submit text extraction:
python scripts/submit_text_extract.py --url "https://www.douyin.com/jingxuan?modal_id=7654800977366715698"
- Poll until
status_messageindicates extraction is complete:
python scripts/poll_text_extract.py --task-id <extract-task-id>
- Return the extracted
textto the user as the source script. Do not rewrite it yet. - Ask the user to choose one next action:
- Use as-is: use the extracted text directly for video generation.
- User edits: wait for the user to provide the modified script.
- Rewrite: ask for rewrite requirements, then rewrite with the model.
- Only produce a rewritten script after the user explicitly chooses rewrite. When rewriting, do not copy protected wording verbatim; preserve structure, pacing, and selling logic only if the user asks for imitation.
- Confirm the final script before direct full generation unless the user explicitly says to proceed.
Original
- Ask exactly:
请提供视频文案。 - Use the user's video script as provided. Do not require brand knowledge, product facts, audience, tone, or copywriting requirements.
- Edit or rewrite the script only if the user explicitly asks. Confirm the final script before direct full generation unless the user explicitly says to proceed.
Video Configuration
Collect:
desc: ask exactly:请描述你想要的数字人,例如人物类型、年龄感、服装、气质或风格。subtitle:onoroff.
The video generation API currently accepts only text, desc, subtitle, and aspect_ratio. Put all character and styling requirements in desc.
The video aspect ratio is fixed to Portrait. Do not ask the user to choose an aspect ratio.
Do not require body framing, camera distance, or pose details. Continue when the user does not provide them.
Preview Or Direct Generation
Ask whether the user needs a preview:
- Ask exactly:
你需要预览还是直接正式生成? - Preview: label this option only as
预览, then submit with--preview, which sends the fixed text测试. - Direct generation: submit with the confirmed final script through
--text. - Do not append any point, price, or consumption note to the preview option or preview result.
Point Confirmation
Before submitting a full generation task, run scripts/submit_video.py --estimate-only with the exact final script. Read required_points from the result.
Tell the user only:
正式生成预计消耗 {required_points} 积分,是否确认提交?
Do not show the source text length, point formula, divisor, rounding method, arithmetic, or calculation process. Do not mention point consumption when offering or returning a preview.
Do not submit a full generation task until the user explicitly confirms the displayed estimate. The submit_video.py script enforces this with --confirm-points; without confirmation, it returns the estimated points and exits without creating a task.
Preview example:
python scripts/submit_video.py \
--preview \
--desc "商务男主播,稳重专业" \
--subtitle on
Direct generation example:
python scripts/submit_video.py \
--text "爱搜数字人视频生成测试" \
--desc "商务男主播,稳重专业" \
--subtitle on \
--confirm-points <confirmed-points>
Always return the taskId to the user so they can manually poll if an automated poll fails.
Poll Result
Use scripts/poll_video.py when the user wants the result for a known task.
python scripts/poll_video.py --task-id a562665671304e29b1f378b65b5d9cd9
python scripts/poll_video.py --task-id a562665671304e29b1f378b65b5d9cd9 --once
On success, return only the useful result fields to the user:
video_urlpreview_url
If polling times out or fails, return the taskId and explain that it can be retried.
Legacy Minimal Flow
For a user who already provides final text and all video configuration:
- Ensure account authorization.
- Estimate point cost with
scripts/submit_video.py --estimate-only. - Ask the user to confirm the required point spend.
- Submit with
scripts/submit_video.py --confirm-points <confirmed-points>. - Poll with
scripts/poll_video.py.
Authorization
Use scripts/auth.py when the user asks to bind, check, or repair AIDSO authorization.
python scripts/auth.py --start
python scripts/auth.py --complete
python scripts/auth.py --status
--status reports whether a local binding exists, authorization is pending or expired, or no binding exists. It never displays the credential and does not verify server-side validity.
Use this two-phase flow:
- Run
python scripts/auth.py --start. It creates a fresh authorization link and exits immediately. - Return the exact
authorization_urlto the user at once as a clickable link. Tell the user it expires in about 10 minutes. - Do not fetch, pre-open, inspect, or validate the link with a plain HTTP request. The authorization page is a JavaScript application; HTTP 200 does not prove that the request is valid.
- Stop and wait for the user to approve the link in a JavaScript-enabled browser.
- After the user says authorization is complete, run
python scripts/auth.py --completeonce. - If the result is
pending, ask the user to finish approval using the same unexpired link. Do not generate another link. - If the link is expired or the page says the request is invalid, run
--startagain and immediately return only the newly generated link. - After
--completereturnsauthorized, retry the original video or text-extraction command.
Never run a blocking authorization poll while waiting for user interaction. Other scripts may generate the same fresh-link response when no local binding exists; surface that link immediately, then use auth.py --complete after approval.
Credential rules:
- Treat authorization as runtime-only. Installation must not check whether a local binding file already exists.
- Installation summaries must present authorization as a first-use runtime step.
- Select and probe a writable local binding path before generating the authorization link.
- Default to
./.aidso/credentials.json; fall back to the installed skill directory or a temporary runtime path if the working directory is not writable. - Never ask the user to paste a key in chat.
- Never print or summarize the key, even masked, in install or status messages.
- Never ship, preload, or directly import a key through environment variables or package files.
- Read the local binding file only after it was created by the authorization flow.
- Store pending authorization state locally with file mode
0600before writing its contents; do not expose the standalonecodein user-facing output. - When no local binding is available, generate a fresh authorization link, return it immediately, then query by
codeonly after the user confirms approval. - Save the returned credential atomically with file mode
0600and never include its value in command output.
All authorization, text extraction, video generation, and result polling requests use the production AIDSO API at https://api.aidso.com. Do not switch to a test, staging, or custom API base URL.
Resources
manifest.yaml: supplementary capability summary; WorkBuddy reads the frontmatter above for skill metadata.scripts/_aidso.py: shared HTTP, credential, and authorization helpers.scripts/auth.py: account binding and status check.scripts/submit_text_extract.py: create a benchmark-video text extraction task.scripts/poll_text_extract.py: poll text extraction and return extractedtext.scripts/submit_video.py: create a video generation task.scripts/poll_video.py: poll a task forvideo_urlandpreview_url.
Read the API details, payload conventions, and examples when verifying endpoint behavior or payloads:
@references/api-spec.md
- Downloads
- Security audit
- Review
- Last updated
- 1w ago
- Current version
- v1.0.1
- License
- MIT-0
