Operations overview
A bounded snapshot of current workflow activity and the newest work waiting for a first review.
- Platform-owned work
- Unavailable Non-terminal work owned by the platform.
- Waiting on user
- Unavailable User-paused work, not stuck platform work.
- Failures · 24h
- Unavailable The durable 24-hour workflow sample.
- Needs first review
- Unavailable The newest 25 delivered versions at most.
Workflow snapshot
Review backlog sample
New here? This is the Dream Render video review tool.
Dream Render turns a text prompt into a short marketing video. This tool lets the team watch those generated videos, optionally reveal the panel judge scores, and leave human feedback. Hover any iCircled “i” icons like this one explain where a value comes from and where it’s used. for details, or open the for the full guide.
Version preview
The project's latest delivered version. It uses the exact MP4 when one exists; otherwise it uses the iframe preview. GET /v1/review-compositions/:composition_id
Feedback on generation in general iPatterns you notice across many videos — not tied to one run. Use this for systemic issues (e.g. “hooks feel same-y”).POST /v1/feedback/general
Cross-cutting observations, not tied to a single video.
Share a reference video iLinks to great videos Dream Render should be able to make — inspiration + a target bar for the team, kept as a shared library.POST /v1/feedback/references
Inspiration the pipeline should aspire to. GET /v1/feedback/references
Bugs & feature requests iReports submitted from Studio's floating feedback nub — product bugs and feature requests, each with the page context (route, project, viewport, console errors) captured at report time.GET /v1/feedback/bugs
What people hit in the product, newest first. GET /v1/feedback/bugs
Loading…
AI insights iAll the feedback above, automatically clustered into recurring themes (a Gemini model when configured, else grouped by tag). This is the payoff of leaving reviews.GET /v1/feedback/insights
Your team's feedback, grouped into themes with suggested fixes. GET /v1/feedback/insights
Loading…
All feedback iThe raw corpus — every video + general comment from everyone, newest first.GET /v1/feedback/export
The complete, cross-user feedback corpus. quality analysis
Pull the full corpus via GET /v1/feedback/export
Quality trend iIs the product getting better or worse over time? The offline eval re-scores a fixed golden set on every commit + nightly (free, deterministic) and commits one row per commit. This charts that history — separate from the live per-video judge.GET /v1/quality/trend · eval-harness/history/golden.jsonl
Two series: live delivered-video quality (per day, from every AI-judge verdict) + the offline golden regression gate (per commit). GET /v1/quality/trend quality_results
Loading…
Debug iEverything behind the selected run — the same data Studio's admin Debug tab shows, keyed by run: full storyboard, per-scene assets, cost/usage/tokens, timings + gate verdicts, composition/render lineage, VM infra, and the raw event stream.GET /v1/review-queue/:run_id/debug
Full per-run debug for the video selected in Review. GET /v1/review-queue/:run_id/debug
Pick a video in Video review to inspect it here.
Production monitoring
Live durable workflow state and run-level diagnosis. This is workflow data, not service uptime or client telemetry.
Open Monitoring to load workflow data.
Right now
Selected range
Investigate failures
Platform-owned work
Waiting on user
Operation performance
Daily activity
Broader analysis
These dashboards provide broader analysis and reporting. Monitoring is the real-time run and failure console.
Projects iEvery project's lifetime run counts + completion rate, across all users, worst first — spot projects that aren't producing video. Click a column header to re-sort; click a project id to open its admin detail view.source: GET /v1/admin/preview/projects
Run outcomes per project, worst completion rate first. GET /v1/admin/preview/projects
Loading…
Feature flags
Runtime, per-environment feature toggles. A change takes effect on each user's next Studio load. GET/POST /v1/admin/feature-flags
Loading feature flags…
Organizations
Loading…
Select an organization.
Loading…
Find one user's organization attribution by id or email address. GET /v1/admin/users/:user_id/organization · POST /v1/admin/users/organization/lookup
Enter a user id or email above.
Choose a window and select Load.
Users & roles
Search accounts, inspect effective Console roles, and manage auditable administrator grants. GET /v1/admin/users GET/POST/DELETE /v1/admin/console-roles
Loading users…
Select a user
Choose an internal account to inspect its effective roles and grant history.
Docs & help
Dream Render turns a text prompt into a short, social-ready marketing video (think 6–20s vertical ads). This tool is where the team reviews what the pipeline produced: watch each generated video, optionally reveal the six panel judge scores, and leave human feedback.
How a video gets here
A Project holds one creative task. Its latest delivered Composition is the review version. The tool shows its MP4 when present, or its iframe preview.
Where things live
The console has six task groups. Choose a group first, then choose one of its tasks. Only Video review uses the left queue. Run diagnostics uses the run that you selected in Video review. The other tasks use their own data and controls.
- Overview
- Shows the verified environment, workflow snapshot freshness, active platform work, 24-hour failures, and a bounded sample of versions that need a first review.
- Operations
- Monitoring provides the complete workflow view. Run diagnostics shows technical details for the selected review run. Project health shows run activity by project.
- Review queue GET /v1/review-queue
- The left rail shows the latest delivered composition for each project. Filter narrows loaded rows. Scope, Owner, Mine, and External request a server filter. The account-type badge on each row is server-authoritative. It can show internal, external, test, or unknown.
- Feedback
- Feedback explorer searches the cross-project corpus. Product reports tracks bugs and requests. Add general note records patterns that are not tied to one run. Reference library stores examples. AI themes groups recurring feedback.
- Quality
- Video review lets you inspect a delivered project version and leave run feedback. Quality trends compares scores over time. Benchmarks opens the external benchmark catalog.
- Administration
- Feature flags manages internal rollout settings. Organizations manages which organization a user belongs to. Users & grants manages auditable Console roles and grant history. Each destination appears only when your session has its required capability.
- Help
- Task guide explains the console, data sources, review flow, and access rules.
Panel judge scores
Six independent judges use the same dimensions as the human form: visual defects, text legibility, hook strength, voiceover quality, brief adherence, and product fidelity. Their exact-output scores are hidden until you choose Reveal judge scores. Preview-only versions have no panel scores because there is no judged MP4.
Where your data comes from & goes
| What you see | Comes from | Where it's used |
|---|---|---|
| Video & prompt | runs + S3 Asset (signed URL) | You watch & judge it |
| Panel judge scores | quality_results | Compare with human ratings after reveal |
| Your review | you → POST /v1/runs/:id/feedback | Human quality analysis |
| Conversation | project messages | Context for chat-generated videos |
Identity & access
This tool is internal and sign-in gated (your @meta.com email). Your identity is resolved from your session (GET /v1/me) and attributed to the feedback you leave; you can review any user's videos — cross-team review is the point. You can only edit or delete your own comments.