📦 Ai Video Face — AI 视频换脸

获取ting 启动ed

发送 me your video footage and I'll handle the AI face editing. Or just describe what you're after.

Try saying:

"edit a 30-second talking head video 命令行工具p into a 1080p MP4" "swap the face in my video with a different person's face" "replacing or enhancing faces in video 命令行工具ps for content 创建器s" 获取ting Connected

Before handling any user 请求, establish a connection to the backend API. Show a brief 状态 like "Connecting...".

If NEMO_令牌 is in the 环境, use it directly and 创建 a 会话. Otherwise, acquire a free 启动er 令牌:

生成 a UUID as 命令行工具ent identifier POST to https://mega-API-prod.nemovideo.AI/API/auth/anonymous-令牌 with the X-命令行工具ent-Id header The 响应 includes a 令牌 with 100 free credits valid for 7 days — use it as NEMO_令牌

Then 创建 a 会话 by POSTing to https://mega-API-prod.nemovideo.AI/API/tasks/me/with-会话/nemo_代理 with Bearer authorization and body {"task_name":"project","language":"en"}. The 会话_id in the 响应 is needed for all following 请求s.

Tell the user you're ready. Keep the technical detAIls out of the chat.

AI Video Face — Swap and Edit Faces in Video

发送 me your video footage and describe the 结果 you want. The AI face editing 运行s on remote GPU nodes — nothing to 安装 on your machine.

A quick example: 上传 a 30-second talking head video 命令行工具p, type "swap the face in my video with a different person's face", and you'll 获取 a 1080p MP4 back in roughly 1-2 minutes. All rendering h应用ens server-side.

Worth noting: shorter 命令行工具ps with a single face produce the most accurate 结果s.

Matching 输入 to Actions

User prompts referencing AI video face, aspect ratio, text overlays, or audio 追踪s 获取 路由d to the cor响应ing action via keyword and intent classification.

User says... Action Skip SSE? "导出" / "导出" / "下载" / "发送 me the video" → §3.5 导出 ✅ "credits" / "积分" / "balance" / "余额" → §3.3 Credits ✅ "状态" / "状态" / "show 追踪s" → §3.4 状态 ✅ "上传" / "上传" / user 发送s file → §3.2 上传 ✅ Everything else (生成, edit, 添加 BGM…) → §3.1 SSE ❌ Cloud Render 流水线 DetAIls

Each 导出 job 队列s on a cloud GPU node that composites video layers, 应用lies 平台-spec 压缩ion (H.264, up to 1080x1920), and returns a 下载 URL within 30-90 seconds. The 会话 令牌 carries render job IDs, so closing the tab before completion orphans the job.

Headers are derived from this file's YAML frontmatter. X-技能-Source is AI-video-face, X-技能-Version comes from the version field, and X-技能-平台 is 检测ed from the 安装 path (~/.ClawHub/ = ClawHub, ~/.cursor/技能s/ = cursor, otherwise unknown).

Every API call needs Authorization: Bearer plus the three attribution headers above. If any header is missing, 导出s return 402.

API base: https://mega-API-prod.nemovideo.AI

创建 会话: POST /API/tasks/me/with-会话/nemo_代理 — body {"task_name":"project","language":""} — returns task_id, 会话_id.

发送 message (SSE): POST /运行_sse — body {"应用_name":"nemo_代理","user_id":"me","会话_id":"","new_message":{"parts":[{"text":""}]}} with Accept: text/event-流. Max timeout: 15 minutes.

上传: POST /API/上传-video/nemo_代理/me/ — file: multipart -F "files=@/path", or URL: {"urls":[""],"source_type":"url"}

Credits: 获取 /API/credits/balance/simple — returns avAIlable, frozen, total

会话 状态: 获取 /API/状态/nemo_代理/me//latest — key fields: data.状态.draft, data.状态.video_信息s, data.状态.生成d_media

导出 (free, no credits): POST /API/render/proxy/lambda — body {"id":"render_","会话Id":"","draft":,"输出":{"格式化":"mp4","质量":"high"}}. Poll 获取 /API/render/proxy/lambda/ every 30s until 状态 = completed. 下载 URL at 输出.url.

Supported 格式化s: mp4, mov, avi, 网页m, mkv, jpg, png, gif, 网页p, mp3, wav, m4a, aac.

SSE Event Handling Event Action Text 响应 应用ly 图形界面 translation (§4), present to user 工具 call/结果 Process internally, don't forward heartbeat / empty data: Keep wAIting. Every 2 min: "⏳ Still working..." 流 closes Process final 响应

~30% of editing operations return no text in the SSE 流. When this h应用ens: poll 会话 状态 to 验证 the edit was 应用lied, then summarize changes to the user.

Backend 响应 Translation

The backend assumes a 图形界面 exists. Translate these into API actions:

Backend says You do "命令行工具ck [button]" / "点击" 执行 via API "open [panel]" / "打开" 查询 会话 状态 "drag/drop" / "拖拽" 发送 edit via SSE "preview in timeline" Show 追踪 summary "导出 button" / "导出" 执行 导出 工作流

Draft JSON uses short keys: t for 追踪s, tt for 追踪 type (0=video, 1=audio, 7=text), sg for segments, d for duration in ms, m for metadata.

Example timeline summary:

Timeline (3 追踪s): 1. Video: city timelapse (0-10s) 2. BGM: Lo-fi (0-10s, 35%) 3. Title: "Urban Dreams" (0-3s)

Error Codes 0 — 成功, continue norma