Built for AI filmmakers who think in shots, not sliders. The flexible core flow is Prepare → Image → Video → Vault: plan the shot and identity, approve a first image, animate that exact image, then save the reviewed recipe and lineage. Finish approved downloads with upscaling and interpolation when needed.
See the workflow →The core apps support one flexible route: Prepare → Image → Video → Vault. Prepare can start from writing, a frame, identity panels, or a prompt; approve the image before it becomes video; then save the recipe. CastLoop media stays available to download, while the current Vault handoff stores recipe and lineage rather than media attachments.
Paste your script, scene, or beat sheet and get a complete shot list with per-shot AI video prompts. Control pacing density, camera style, and add director notes. Outputs ready-to-use prompts for Kling, Runway, Veo, and Pika — one click to copy or download.
A model cannot keep a face it has never seen. Forge generates the hero image and every reference panel from it — clean single-subject images a video model can take as input. Build makes that a record: the identity spine, the traits that must never drift, wardrobe as versions rather than edits. The LoRA lab trains a Consistency Model from those same panels and checks the model renders the same person. Sequence then attaches the record to a shot list, so shot 40 describes the same person as shot 1.
Full cinematographic control over every parameter — camera body, lens, aperture, shot type, movement, lighting, color grade, film stock. Receives input from FrameDNA or start from scratch. Outputs optimised prompts for Kling, Runway, Veo, and Midjourney V8.2.
Store the prompts, references, characters, seeds, and notes that actually worked. DVP Vault is the recipe-and-lineage memory layer; the current CastLoop save does not claim durable media attachments.
Once shots are generated and approved, upscale them from preview-quality to delivery-ready. SeedVR2 and Topaz upscaling at 2× or 4×, Topaz Precision with face enhancement for character work, and Topaz Interpolate to take the frame rate to 60fps. Runs on fal.ai with your own API key.
The connected workflow page for VibeBoard, Campaign Engine, and DVP Vault: define the creative direction, ship the campaign, then recall or fork the memory later.
AI creative direction before you shoot, pitch, or publish. Capture the idea, generate a Vibe DNA and board with OpenAI or Gemini (or a local draft), score new shots for consistency, then hand off a packet straight to Campaign Engine or publish it to DVP Vault.
A browser memory layer for approved prompts, references, source packets, and Campaign Engine packs, plus CineQC QC records and reusable Shot Doctor fixes. It remembers what worked, so the same fix is a recall next time.
Have a reference frame already? FrameDNA reads its cinematography — shot type, lens, lighting, colour grade, film texture, mood — and loads it straight into CinePrompt. One way to fill in a prompt, not a step every project takes.
One Cinema Desktop for character continuity: Forge clean reference panels a model can take as input, Build them into a versioned record through Library → Character → Inspector, train a Consistency Model from those panels in the LoRA lab, then run Shots → Edit → Output so identity, wardrobe, lighting and expression inherit shot to shot. Drafts autosave locally, with optional account cloud sync.
Compare the major AI video platforms side by side by quality, features, audio, speed, and cost so you know where each shot should actually be generated.
Author reusable Camera Motion DNA with measurable movement, three-phase timing, deterministic schematic previews, and portable model-neutral JSON.
Paste a failed AI video prompt and what went wrong, then get a diagnosis, a repaired prompt, a better model route, and a Vault-ready repair note. Optional AI visual diagnosis of a failed still or short clip using your own key.
Estimate spend across shots, platforms, durations, and iteration counts before production starts so you can choose the right workflow for the budget.
Type a natural-language description and get a model-aware image prompt — Midjourney flags, FLUX/SDXL prose + tags, or structured JSON for Imagen/Gemini — with auto-detection and full camera/quality overrides. Now the Stills / JSON tab inside CinePrompt.
Design any voice in your browser: independent pitch and formant control, breath and rasp texture, tremble, creature layers, character macros, a four-corner Morph Pad, and a describe-it Character Builder. Optional AI target voices with your own ElevenLabs key, or generate a take from typed text with a Fish Audio key.
Drop any audio or video file and get a formatted .srt subtitle file in seconds. Powered by OpenAI Whisper running entirely on your machine — fully offline after first download.
Paste client or producer review notes with timecodes and export import-ready marker files for DaVinci Resolve, Premiere Pro, Avid, and Final Cut. Offset to your timeline start, entirely in the browser.
Dial in five dimensions of your shot — motion, dialogue, style, complexity, importance — and get ranked AI platform recommendations with the reasoning behind every call.
Check browser video with waveform and vectorscope tools when you need a quick technical pass on color and signal levels while reviewing renders.
Run broadcast delivery checks on AI video: EBU R128 and other profiles with true peak, loudness range, safe areas, captions and HDR metadata, flash and levels screening, a scored rubric, frame-by-frame CineQC Diff, a C2PA Content Credentials check on every file, an AI artifact explainer, and PDF reports with a share link. Analysis runs in your browser.
Design evaluation rubrics and run blinded AI-video evaluations locally: score outputs, confirm preference rankings, and export defensible training or preference data as JSONL. Media never leaves your browser.
Part of CineQC. Verify C2PA Content Credentials in your browser: read the full provenance chain, check signers against CAI anchors or your own PEM list, detect stripping, and export JSON or PDF reports. CineQC runs the same check on every clip it scores.
Build a campaign brief, pull source material from DVP Vault, generate or hand off AI drafts, generate AI backgrounds (Imagen 3 / Gemini Flash Image), edit every slide, check safe areas, then export platform-ready social and content packs.
Match display-referred stills to a reference look, preview the exported grade, and download a creative .cube LUT for a matching color-managed pipeline.
Generate clean digital production slates for your shots — scene, take, roll, and metadata — ready to drop on frames or hand to the edit.
Foley & soundscape prompt builder for AI audio models — craft precise prompts for ElevenLabs SFX, Stable Audio 2, MMAudio, and AudioLDM, or render the SFX in-browser via fal.ai with your own key.
Generate original score with Google Lyria — direct the mood, genre, tempo, and instrumentation to soundtrack your scene without licensing headaches.
Analyze song key, BPM, Camelot code, waveform, and loudness locally in your browser, with prompt-ready music direction for matching cuts to track.
The edit bay's AI sound intelligence — retime, beat-sync, and split stems so you can cut picture to music and tighten the rhythm of a sequence.
Drop in a generated clip and a separate music or voice track, scrub them on a shared timeline, mark the beat hits and cut points, and export a timing sheet — before you ever open an NLE.
Hands-on evaluation of the major AI video models — how they actually perform on real shots, with the strengths and failure modes behind the marketing.