ENGINE ยท 04 ยท CHANGELOG

Shipped recently.

Every release, in the order it landed.

SEP 20, 2026

Watch the cut before you pay to animate it

  • โ†’The Stills gate now plays an animatic: your frames against the real voiceover, at the real timings, with the same Ken Burns moves and the same 0.6 second dissolves the renderer uses.
  • โ†’Beats you have not drawn yet are held on a labelled card for their true length rather than skipped, so the rhythm on screen is the rhythm of the finished video even before every frame exists.
  • โ†’It is required. Stills cannot be approved until the cut has been played through, or until you say you have seen enough - and you can only say that once you have taken the playhead to the end. Clips is the stage that spends, and this is the last free look at the timing.
  • โ†’Nothing is generated by opening it or playing it. The same player is still on Deliver, where it is the free rehearsal before a paid render.
SEP 10, 2026

Set up a video from one prompt

  • โ†’Describe your video at the top of New video and Cue fills in the source, brief, title, format, target length, and generation tier for review.
  • โ†’Your playlist supplies the Look, voice, and brand defaults. Everything remains editable, and generating the setup does not create or render a video.
SEP 06, 2026

AI video news - a public destination for the daily briefing

  • โ†’/news lists every published edition and /news/{date} renders one, with each story linking out to the publication that reported it.
  • โ†’The stream was re-aimed at the model set we rank on /best-ai. Queries are built from the tracked model, vendor, platform and policy terms rather than the broad technology, entertainment and business news buckets it used to sweep.
  • โ†’Attribution is joined back to the fetched article by index and never read from the editor model, so a rewritten story cannot be filed under a publication that did not write it. A pick that cannot be traced to a source is dropped.
  • โ†’An edition is published only when curation finishes with at least three stories that clear the relevance bar. Thin editions stay out of the page, the feeds and the sitemap alike.
  • โ†’RSS at /feeds/news.rss and JSON Feed 1.1 at /feeds/news.json, in the same envelopes the channel and playlist feeds use.
AUG 04, 2026

Song to character music video

  • โ†’Upload a song on the new "From a song" card in the video chooser: MP3, WAV, M4A, AAC, or FLAC up to 100 MB.
  • โ†’Lyrics are transcribed and time-aligned to the vocal. Paste the lyrics yourself and we force-align the text instead of guessing at sung vocals, which gives noticeably tighter cuts.
  • โ†’One shot per lyric line, with instrumental intros, solos, and outros subdivided on the track's own energy peaks.
  • โ†’Your Characters and your Look drive the visuals; stills and motion clips follow the existing motion-clip percentage. No lip sync in v1.
  • โ†’The uploaded song is the entire audio bus: no narration, no ducking, no looping. A rights acknowledgement is required at upload: publishing a render of someone else's recording to YouTube will draw a Content ID claim.
JUN 04, 2026

Creator Library v1 - Characters, custom Looks, and the Mine pattern

  • โ†’Characters: new first-class element type. Build a recurring presenter once - references, voice, personality, base appearance - and cast them across every video and every channel.
  • โ†’Custom Looks: fork any catalog Look or build from scratch; per-user authoring lives under the Mine tab on `/app/looks`.
  • โ†’Voices Mine: Library = Voices now ships a Catalog | Mine tab pair; ElevenLabs voice clones surface as cards alongside the curated catalog.
  • โ†’Render pipeline integration: a `cast` block sits next to `look` in project.json; the image prompt composes character + Look + scene so identity stays consistent across aesthetics.
  • โ†’Image and Video LoRA training: per-character premium rails on Studio and Portfolio. Train once, lock identity, switch a per-render dial between Ref and LoRA.
  • โ†’Four new guides on `/guides`: Build your first character, Author a custom Look, Train a character LoRA, and Ref vs LoRA picking.
MAY 20, 2026

Portfolio operator suite - the channel-of-channels wave

  • โ†’Topic Queue: bulk-import, source-grounded scripts, idea โ†’ research โ†’ produced pipeline across every channel in a portfolio.
  • โ†’Source Packs: primary + secondary citations attached to claim-heavy topics. The script generator quotes your sources, not hallucinations.
  • โ†’Companion Articles: every video auto-publishes a matching SEO page with transcript, chapters, FAQ schema, VideoObject JSON-LD.
  • โ†’YouTube Compliance Mode: private-first uploads, audit-readiness, API quota tracker. Pre-positions you for the API audit.
  • โ†’Comment Inbox: read comments across every connected channel; Cue drafts replies; you approve. Never automated.
  • โ†’Series Archetype landings: six niches with starter topic packs - How Things Work, Business Breakdowns, AI Operator Playbooks, History Visualized, Travel, Finca / Property / Lifestyle.
  • โ†’Niche Scorer and YouTube Audit: two free public tools that funnel into the Portfolio plan.
MAY 15, 2026

Marketing surfaces - solutions, case study, free tools

  • โ†’/solutions hub with nine industry playbooks (creators, SaaS, news, e-commerce, education, agencies, local business, developers, portfolio operators).
  • โ†’/case-studies/fincabout - long-form portfolio operator case study.
  • โ†’/best-ai - head-to-head AI model rankings across six categories.
  • โ†’/tools index expanded to 27 free creator utilities.
MAY 08, 2026

Shorts workbench + multi-platform publishing

  • โ†’Auto-detected 9:16 clips from long-form uploads - approve, edit, push to TikTok / Reels / Shorts.
  • โ†’Per-channel publishing destinations with OAuth - YouTube + TikTok + Instagram.
  • โ†’F2 free-form text overlays + F3 per-cue color grade + F10 auto-duck on music tracks.
APR 30, 2026

Cue + slash commands + per-section "Ask Cue"

  • โ†’/tighten, /broll, /rewrite, /shot-list, /lock, /voice - six verbs that do the work.
  • โ†’Per-section button on every editor block - Cue answers in the context of that exact slice of the cut.
  • โ†’Mood-board, narrative arc, B-roll prompt generation routed through Cue.
APR 22, 2026

Frame-accurate captions + voice clones

  • โ†’Whisper-based caption pipeline with editable cue list per video.
  • โ†’ElevenLabs voice cloning on Creator plan and up.
  • โ†’SRT / VTT export plus burned-in caption styles.
APR 10, 2026

Templates v1 - 35 cinematic channel formats

  • โ†’/templates catalog with poster-SVG previews and per-template landings.
  • โ†’Brief-in / library-out pipeline for repeat publishing cadence.
  • โ†’Per-template recommended voice, music mood, render cost estimate.
MAR 28, 2026

Foundations - auth, billing, render pipeline

  • โ†’Google auth, Stripe billing, plan tiers, per-key API access.
  • โ†’Remotion render pipeline with queued retries and progress updates.
  • โ†’Filmwiki seeded with 249 film terms across six taxonomies.

Your first cut
is on us.

+ New videoFree plan ยท no card ยท 1080p export