course-video-studio

Turn finished lecture scripts into presenter-led course videos with lecture narration. This course video studio prepares the spoken narration for each lesson, then records a digital-teacher delivery, so enablement teams get ready-to-publish lecture videos and training presenter clips.

Turn finished lecture scripts and one authorized teacher portrait into lesson videos. Each lesson is narration first, then talking-head clips that fully contain the speech.

Scope and adjacent routes

Use this Skill when a course or training team wants talking-head lesson videos from a stable portrait plus lecture text.

Route audio-only course work to course-narration-studio. Route a single presenter clip to talking-avatar-video. Route a founder weekly update to founder-ip-avatar-studio. Without authorization for the likeness and the voice, stop at confirmation.

When the teacher has a slide deck (PowerPoint, WPS, or PDF) and wants the videos made from the slides themselves, point to Lecture Video Maker (lecture-video-maker) for a whole course or Micro-Lesson Maker (micro-lesson-maker) for one lesson; when there are no slides yet, AI Courseware Maker (ai-courseware-maker) builds the deck first. Use the install message in Course video workflow and install it only after the user asks. Talking-head lessons from a portrait and final lecture text stay here.

Inputs and defaults

The hard inputs are:

• one accessible teacher portrait;
• final lecture text for at least the pilot lesson;
• lesson order;
• likeness and voice rights;
• a pronunciation table whenever the lecture contains recurring names, terms, or acronyms.

If those terms exist and the table is empty, stop and collect the readings. Do not guess them. Final lecture text is the spoken source. A deck is useful only after the user extracts the spoken script; this Skill does not parse slides.

Reuse the frozen voice_id, language, and framing. Default to model: "auto" for video and a source-derived aspect ratio. Split a lesson into sequential containable clips when planned speech is longer than the longest video duration that can hold it.

Golden path

1. Inspect the portrait. Record MIME type, width, height, aspect ratio, byte size, and whether it has an alpha channel. Build the lesson ledger. For a local portrait or sample, upload only through the bundled client after inspection. Never pass a local path to a remote tool.

2. Select or clone a teacher voice. A clone requires consent first; access to a file is not consent. A suitable sample is currently about 10 to 300 seconds, no larger than 20 MiB, and clean single-speaker speech. Use the live voice_clone admission card before any clone client_request_id or beatra.voices.clone call: route voice_clone, tool beatra.voices.clone, live price for one successful activated voice, the provisional estimate, the 600-credit signup gift usually cannot start this clone, and what happens if the balance is short. Do not create client_request_id or submit until the user confirms they have topped up or already have enough credits. Do not offer a free clone. Submit beatra.voices.clone exactly once:

   json

   {
     "sample": {"type": "artifact", "artifact_id": "artifact_..."},
     "display_name": "Course Teacher",
     "consent_attested": true,
     "model": "auto",
     "client_request_id": "opaque-course-video-clone"
   }
   

   Poll with beatra.tasks.get until terminal, then freeze the returned voice_id. Without a clone, call beatra.voices.list when a catalog voice still needs to be selected. Never put a display name in voice.

3. Call beatra.models.list for text_to_speech and image_to_video. Confirm the video route admits [image, driving_audio]. Compare the portrait's actual MIME, dimensions, byte size, and alpha-channel presence with every advertised image constraint. Read the longest video duration that can fully contain speech. Split the pilot lecture into sequential containable segments at sentence or section boundaries. Do not shorten the lecture to cheapen a clip. Do not ship a truncated lesson. Use mp3 only when the live speech card supports it and the live video card accepts audio/mpeg. If the user requested flac, opus, or pcm and the video route does not accept the resulting format, explain the incompatibility before any paid call and obtain the user's choice. Do not silently change the format. If any required media fact is unavailable or incompatible, stop before TTS. Keep each speech input at or below 50,000 characters.

4. Confirm the pilot speech card: script segments, voice, format, live estimate, and one client_request_id per beatra.speech.synthesize call. Submit each segment once through the bundled client. Poll each narration task with beatra.tasks.get until terminal. Read the returned artifact plus actual task.output.audio.mime_type, duration_seconds, and size_bytes. Approve the real audio before that segment's video. Never treat a script preview as an audio review.

5. Refresh the image_to_video cards and admit the actual portrait plus that segment's approved speech. Recheck every image and driving-audio fact. If terminal audio size remains unavailable, stop before video. Use the smallest supported integer video duration at or above the actual speech length. Show the video admission card before beatra.videos.animate: route image_to_video, tool beatra.videos.animate, approved portrait and speech artifacts, prompt, audio-led duration, resolution if set, output count, provisional live estimate, the 600-credit gift usually cannot start this video, and what happens if the balance is short. Do not submit until the user confirms they have topped up or already have enough credits. Approved narration does not authorize the video call. Each segment video has its own client_request_id.

6. After the pilot lesson's clips are accepted, repeat segment-narration-then-video for remaining lessons. Each paid stage has its own client_request_id. Submit only through the bundled scripts/mcp_client.py. Do not configure a host Beatra Connector. Do not use REST/OpenAPI as a fallback. Poll every video task with beatra.tasks.get until terminal.

7. Deliver labeled lesson videos in teaching order, with segment order inside a lesson, plus actual dimensions, duration, usage, and billing.net_charged_credits. Review identity, term readings, and lip timing.

Decisions that require confirmation

Confirm rights, the voice, the pilot, and every video admission card. A changed lecture is new narration and, if the audio changes, a new video.

Recovery

Recover each stage only with its identical frozen payload and ID. Recover a lost task ID through beatra.tasks.list and beatra.tasks.get. On insufficient_balance, keep the top-up URL inside the balance error exact and retry the same frozen client_request_id only after the user says they have topped up.

Account balance

When the user asks how many credits remain or whether a live estimate fits, call beatra.wallet.get. When they ask what was charged, call beatra.wallet.ledger. Both are read-only. Do not invent an account-balance or top-up tool. Do not make wallet.get a required step before every paid submit.

When a model card comes back carrying a top_up block, relay its tiers as the card lists them and in that order. Do not rank them, do not talk one down, and do not pick one for the user. Which tier suits them is their call, made on the wallet page with the whole list in front of them. Never quote a tier from memory.

References by task

• Read Course video workflow for ledgers, admission cards, payloads, and review.
• Read Installation and authentication only when authorization needs attention.
• Read Installation registration for first-use registration.
• Read Tasks and results and Billing, errors, and recovery for task and billing facts.
• Read Bundled MCP Client diagnostics when the bundled client cannot connect.
• Updates and security: updates and security
• Read Uninstall and disconnect only when the user asks to remove the package.

Use or invoke the bundled scripts/mcp_client.py for every Beatra operation.

Updates and security

This Skill runs only the files installed with it and leaves them unchanged while it runs; it never downloads or executes other code. It talks to Beatra's official services over HTTPS, and a file you choose to upload goes only to the one-time upload address Beatra issues for it.

Once a day at most, the bundled client reads Beatra's public version list. When the store or tool you installed from has a newer version, it prints one line with the command that updates it there. Updating stays your choice, through that store or tool. See updates and security.