TARKOVSKY VS MORPHIC
Tarkovsky
vs Morphic.
Morphic is the closest in shape: an AI video and image studio that routes to other makers' models, with a canvas, a chat copilot and an agentic editor. Tarkovsky routes to other makers' models too. The difference is what each one asks you to hold: Morphic a canvas of generations, Tarkovsky a film's plan.
Reviewed 27 October 2026, from Morphic's own pages (linked under Sources) and Tarkovsky's field guide (Tarkovsky 0.46). Prices and features change; check their pages before you buy, and tell us what we got wrong.
STAGE BY STAGE
What each one does.
| Stage | Morphic | Tarkovsky |
|---|---|---|
| What it is | “Morphic is an AI video and image studio — from ideation to a finished story.” A browser studio that aggregates dozens of third-party image, video and audio models behind a canvas, a chat Copilot, saved references, node workflows and the Compose editor, paid in credits. | The independent AI film studio: one project holds the script, the characters and places, the shot list, the storyboard, the takes, the sound, the cut and the export, and each stage reads the one before. |
| Script and story | Write with Copilot: it breaks down scenes, refines dialogue and explores alternate arcs. Copilot's replies cost credits by model. No screenplay editor or script lock is described. | A real screenplay editor (Enter and Tab as in a screenwriting app; Final Draft, Fountain, PDF, Word and plain text in and out), a free rule-based script check, a chat that drafts and changes the script with the language model you choose, and Lock, which makes the scenes and a sheet for every character, place and thing. |
| Characters and places | Character and style references from a few images or one, passed to the chosen model (nothing is trained); a character sheet per recurring character and a location sheet per recurring place. One saved reference on Free. | One identity picture per character and every angle made from it; a look per costume or age; a floor plan drawn at Lock for every place, free, with plates for each side and time of day. Every image is checked as it lands, and the heart marks the one every shot is sent. |
| Storyboard | A cinematic storyboard workflow with shot angles and mood, and a 3×3 storyboard grid. | A still for every key frame of every shot, made with the characters' and the place's references and checked as it lands; a wall of frames arranged in the order they cut; board versions saved at Lock; an animatic. The still you love is the frame the take starts from. |
| Shot planning and direction | Direct scenes and iterate on shots on the Canvas; camera motion, 3D motion, frames to video, inpainting and outpainting. No shot list, floor plan or 3D viewfinder is described. | A Shot List by beats: size, lens, move, length, focus, staging, screen direction and a lighting plan per scene, drafted by Propose coverage. A 3D Viewfinder built from the floor plan. A directing studio where cues (gaze, move, camera, face, line, tempo) sit on the shot's timeline and say how strongly they reach the video model. |
| Video generation and models | No model of its own. MiniMax H3 Max, Seedance 2.5 (the default for most video work), FLUX 3 Max, Wan 3.0, Veo 3.1, Kling 3.0, LTX 2.5 and more; text, image, frames and reference to video, lip sync, sound effects for video. | No model of its own. Veo 3.1, Kling 3.0, Seedance 2.5, Grok Imagine Video 1.5, Wan 3.0, MiniMax Hailuo and more, chosen per project or per take; each model's line says its sound, the references and frames our route sends, and its price a second. Every take is checked as it lands for a cut inside it; a motion check and a text check run on a button, or on every take if you choose. |
| Editing | Compose, “the agentic video editor”: prompts become edits on a real multi-track timeline (trim, reorder, retime, reframe, grade, transitions, titles, captions, animation, mixing), with dialogue, music and effects lanes, fades and ducking. | A nonlinear timeline: sequences, picture tracks, tracks for dialogue, music, effects, ambience and narration, trims and rolls with snapping, transitions, titles, captions, end credits and cutdowns. The Assembly builds the first cut from the loved takes, free; the Rough Cut proposes one beat by beat; a clip can be remade from the cut. |
| Sound, music and voices | Speech, dialogue, music, sound effects, voice change and transcription through the routed models (ElevenLabs, Fish Audio, Lyria, Seed Audio and others), paid per model in credits. | Music cues composed in sections that turn on the cuts (ElevenLabs Music), effects and ambience made from a marked range (ElevenLabs Sound Effects), Kokoro or ElevenLabs voices; on Tarkovsky credits or your own ElevenLabs key. Placed sounds play in the preview and the render. |
| Export and hand-off to an editor | MP4 in H.264 or H.265 from 720p to 4K, captions burned in or as SRT/ASS; rendering in the browser is free, on Morphic's servers it costs credits. No export to Premiere, DaVinci Resolve or Final Cut is mentioned (only a keyboard-shortcut preset that follows them). | MP4 (H.264) or ProRes at 720p or 1080p with a look, a LUT, grain, halation and burned-in captions or timecode; the project to DaVinci Resolve and Final Cut Pro (FCPXML) or Premiere Pro and Avid (EDL) with every clip's media. Renders never cost credits. 4K is not rendered yet. |
| The production record | Not described on the pages read. | Every still and take keeps the prompt it was sent, exactly as shown before it ran, its references, its model, its price (reserved at the quote, settled to the provider's receipt) and what the checks found. Every page has a History with Undo; nothing is deleted. |
| Bring your own keys | Not mentioned on any page read: “Generations use Morphic credits.” | Yes. OpenAI, Anthropic, Google Gemini, OpenRouter, MiniMax and ElevenLabs keys, encrypted per account and checked once when saved; the provider bills you and Tarkovsky charges no credits for that call. The makers that may train on what you send are marked before you choose a model. |
| An AI director | Compose edits by prompt and Copilot writes and plans; no agent that makes the whole film within a budget is described. | Tarkovsky Director: an AI model directs and produces the film within a budget you set, through the same routes, checks and prices as the page. The kickoff says what it may decide; it never loves an unchecked still or waves a must-fix flag through, and reports every decision and its cost. |
| Collaboration | Live collaboration: several people in one file, comments anchored to an element, Admin, Editor and Reviewer roles and per-member credit limits on Pro and above. | Owner, Editor and Viewer roles per project: one collaborator per project on Free, three on Creator, ten on Studio, 25 on Production. |
| Pricing (read 27 October 2026) | Free $0: up to 20 credits once, one saved reference. Basic $19 a month: 2,400 credits. Standard $24: 3,625 credits, Canvas, Copilot, Compose, upscaling, commercial use. Pro $45: 6,350 shared credits, up to four more members at $10 each. Max $170: 24,650 shared credits. Education $4: 610 credits. Credit packs from $25 (2,500) to $500 (62,500), valid 180 days. Subscription credits reset monthly and unused ones expire. Yearly prices did not render on the page the day it was read. | Free $0: three projects, 500 welcome credits once, 500 MB. Creator $19 a month: 1,250 credits a month, 25 projects, 10 GB. Studio $49 a month: 4,000 credits a month, 100 projects, 40 GB. Production $199 a month: 20,000 credits a month, 500 projects, 200 GB. Credits on a paid plan don't expire. Top-up packs cost 2¢ a credit on Free, 1.5¢ on Creator, 1.2¢ on Studio and 1.0¢ on Production. One credit is $0.01 of usage; a paid call costs what the provider charges plus 50%, shown before it runs. |
| The free tier | Up to 20 credits once, limited models, workflows, one saved reference, no card required. | Three projects, 500 welcome credits once, every stage and your own keys. The sample film Emina's Crossing is in every account, free to keep. |
| Platform and company | In the browser; Morphic (founder and CEO JD Kanani), offices in San Francisco, Bangalore and Dubai; Morphic Studio launched in July 2024. | In the browser, from Tarkovsky AI; an alpha release (0.46, October 2026), so export anything you care about. |
IN FAIRNESS
Where Morphic is stronger.
Morphic has Compose, an editor that takes a prompt and makes the edit on a real timeline, export up to 4K, a live multi-user canvas, an MCP server and a partner API, and a longer list of model modes (lip sync, inpainting, 3D motion).
THE DIFFERENCE
Where Tarkovsky differs.
Tarkovsky holds the film's plan rather than a canvas: a screenplay editor with Lock, floor plans and a 3D Viewfinder, a shot list by beats, checks on every still and take with the record of what was sent, hand-offs to Resolve, Final Cut, Premiere and Avid, your own keys, Tarkovsky Director within a budget, and credits that don't expire on a paid plan.
SOURCES
What we read, and when.
Morphic's pages, read 27 October 2026:
- Morphic pricing
- Plans and credits (docs)
- Pricing FAQs (docs)
- Filmmaking
- Compose
- Character and style references
- Collaboration (docs)
- MCP server (docs)
- About Morphic
Tarkovsky's column is the field guide as of Tarkovsky 0.46. A cell that says "not described on the pages read" means we found nothing on the pages above, not that the feature doesn't exist.