How do I revise a finished video from written feedback without redoing the parts I already approved?

the agent that answers this

Final Video Editor With Avatar Presenter. Final Video Editor: Take a final video (fully rendered with b-rolls + animations), take user's feedback as plain text/a file, then edit the video to generate the final render. Use a talking-head avatar presenter.

• Step 1tool

  Load the supplied finished video and the user's feedback notes or feedback file, confirm the video exists, extract metadata, and normalize the feedback into concrete edit requests.

  Your model fills this step

  • No integration? If the feedback is pasted as plain text instead of a file, use the pasted text directly.
  • No integration? If timestamps are missing, infer likely edit locations from transcript, scene descriptions, or visual review and flag assumptions in the edit plan.
• Step 2

  Analyze the current video against the normalized feedback, identify which existing b-rolls, animations, audio, captions, timing, or avatar presenter moments need edits, and produce an edit decision list.

  Your model fills this step
• Step 3tool

  Preflight: immediately before the first dependency-using step, make sure the remaining work has what it needs, so the run never fails halfway on a missing tool or an unconfigured account. Services and accounts (these need a key, login, or subscription, NOT auto-installed): Runway ML (usage-based (credits); get it at https://dev.runwayml.com; alternative: Seedance (via HeyGen)); HeyGen (free trial, then paid; get it at https://heygen.com). Before the FIRST step that calls a paid service, verify its API key or login is configured. If one is missing, STOP and tell the user exactly what to sign up for, the link, and any cheaper alternative, instead of failing partway through a paid run.

  Your model fills this step
• Step 4tool

  Generate or revise talking-head avatar presenter clips from the edit decision list using the configured HeyGen avatar, voice, script, framing, and timing requirements.

  Your model fills this step

  • No integration? If HeyGen requires manual browser login or a render is still processing, hold the run as Needs You with the exact HeyGen project/render URL, the expected local destination path, and the next download instruction.
  • No integration? If the feedback does not require any new avatar presenter speech, reuse the existing presenter segment and produce an explicit no-new-avatar note.
• Step 5tool

  Generate or replace b-roll assets in Runway only for feedback items that require new visual footage, animation changes, or visual inserts not already present in the source video.

  Generate or extend videos with Runway through RunAPI. Use the RunAPI CLI for one-off results and an SDK when integrating Runway into an app, backend, worker, library, or production codebase.

  --- name: runway description: "Generate or extend videos with Runway through RunAPI. Use the RunAPI CLI for one-off results and an SDK when integrating Runway into an app, backend, worker, library, or production codebase." documentation:…

  Full skill: runway

  • No integration? If Runway API credentials are unavailable, use signed-in browser access when possible.
  • No integration? If a Runway render is submitted but not yet downloadable, hold the run as Needs You with the exact Runway render URL, the expected local destination path, and the next download instruction.
  • No integration? If no new b-roll is needed, skip this step and preserve the existing b-roll and animations.
• Step 6

  Prepare the local edit project: cut or retime the source video, insert avatar clips, preserve or replace b-rolls and animations according to the edit decision list, update captions, balance audio, and build a render manifest.

  Your model fills this step
• Step 7tool

  Render the revised final MP4 locally from the prepared edit project, using the requested output folder and aspect ratio.

  Your model fills this step

  • No integration? If Remotion is not installed but the edit can be expressed as FFmpeg operations, render with FFmpeg and record the command used.
  • No integration? If required source assets are missing, stop as needs-input naming exactly which asset path is required.
• Step 8

  Run final QA on the rendered video: verify the file exists, plays, has audio, matches requested duration/aspect ratio, applies each feedback item, preserves unchanged creative elements, and includes the avatar presenter correctly.

  Your model fills this step
• Step 9decision

  Quality gate: adversarially critique the deliverable produced by the prior steps, from three lenses (a domain expert, a hard skeptic, and the end user). Check it actually accomplishes the job, is accurate and on-brand, and is genuinely high quality. Fix clear problems in place; if something should block or materially change it, say so before it reaches the user.

  Decision step
• Step 10decision

  Voice Pass: before showing any user-facing text, rewrite the final draft so it sounds natural, platform-native, and non-generic. Preserve the meaning, facts, links, citations, decisions, and concrete details. Remove obvious AI tells: generic praise, inflated significance, rule-of-three padding, em/en dash overuse, chatbot artifacts, fake-candid openers, promotional filler, vague hedging, and formulaic conclusions. Match any provided voice samples, agent memory, prior feedback, and platform norms (Reddit/HN/social/email); if none exist, use a concise natural voice. Do not add unsupported claims. For factual, technical, legal, or reference outputs, keep the voice plain and neutral rather than adding personality.

  Decision step
• Step 11decision

  Hold the final render for user review with the final video path, QA summary, and concise list of edits applied; ask for approval before any posting or external delivery.

  Decision step

common questions

1. v3Aug 20generated

   reconciled active outcome capability input authority

2. v2Aug 20generated

   added immutable outcome-search identity

3. v1Jul 26generated

   auto-generated from "Final Video Editor: Take a final video (fully rendered with b-rolls + animations"

Agents are alive: every change is a version, and a run can propose improvements that get reviewed and applied.

related agents

• on-demandcreator

  How do I generate cinematic B-roll clips from a source video and a shot-direction brief using Runway?

  cinematic b-roll generator

• on demandmarketing

  How do I combine Runway-generated B-roll with a HeyGen talking-head avatar into one finished video?

  Cinematic B-Roll Generator with Avatar Presenter

• video production

  How do I turn a markdown animation brief into working Remotion text-animation components?

  Markdown Brief to Remotion Text Animations

• builder

  How do I remove filler words, false starts, and dead air from a raw video recording automatically?

  Raw recording → clean cut (Claude + Remotion)