AI video editing

Your footage,
edited by AI.

Lumière is the AI video editor that turns hours of raw footage into a story-ordered first cut: transcribed, structured, subtitled and checked against your own guidelines. You review every edit. The final say stays yours.

Free trial with 15 lumens ✦ · Hosted in the EU

Our vision

Every story you filmed deserves to be told.

The story is already there, somewhere in hours of interviews, rushes and recordings. Today it waits for someone with a free night to find it. We believe finding it should take minutes, not nights.

Our mission

The camera stops.
The story starts.

We build the AI that does the grind of editing: logging every word, finding every moment, drafting the cut, checking it against your rules. So the people behind the footage spend their time on the story. The editor stays the director.

How it works

Upload. Write the brief.
Review the cut.

Lumière transcribes every word, can also read what was on screen, reads the format from your brief, and arranges the strongest moments in story order. A draft cut lands on your timeline, ready for your judgement.

Your story. Mastered.

Your footage. One prompt.
A first cut.

In numbers

Figures you can check.

No inflated multipliers. Every number below is something Lumière does today.

8 phases
From raw footage to a rendered master, with B-roll from your own footage and stock.
17 formats it knows
Documentary to TikTok, read from your brief. Pacing applied for you.
4 aspect ratios
One master reframed to 9:16, 1:1, 4:5 or 16:9.
1✦ per source minute
Pays for a full draft through the pipeline. Your edits and re-renders are included.
0 frames trained on
Lumière never trains a model on your footage.
R128 loudness, measured
Every compliance job gets LUFS, true peak and range with a pass or fail.

Produce · The story engine

Tell the story in words. Lumière drafts the cut.

Lumière transcribes every word and, when you switch it on, reads the visuals too, so it knows what was said, about which moment, and what was on screen at the time. It understands the content, not just the soundbites. Say what you’re making in your own words, a one-hour documentary or a 30-second Reel, and it cuts to that format’s conventions. Your brief has the last word.

The footage Three hours of recording. Five people, each telling their part of the same group trip: out of order, over each other, the way people really talk.

Brief
“Open with the preparation. Then the journey and the first evening, the dinner where the surprise guest walked in. Build to the event itself, then the impressions that moved people most. Close on the host’s own words. The best telling of each phase, documentary pacing, one hour.”
Draft cut · 1:00:00 · read as documentary
The draft lands on your timeline. You review every edit; every decision stays yours.

Eight phases, one pass, from upload to a rendered master.

  1. Ingest
  2. Transcribe
  3. Understand
  4. Story
  5. Script
  6. B-roll
  7. Music
  8. Master

Where the pipeline is headed, not sold yet: 360° virtual-camera editing and named-entity memory.

Find

Every word. Searchable.

Somewhere in hour two, the speaker finally says it. Search the transcript for the words you need and jump straight to each moment, or ask the footage a question and get the timestamp back.

6 matches across 3 hours of footage

Built for newsrooms. Coming Topic search across all your rushes: every passage on a subject, in one list.

Journalists · Newspapers · Magazines · Radio

The quotes are already written down.

Record the press conference, the interview or the broadcast, and upload the video or the audio. Lumière writes down every word with its timestamp, sums it up in chapters and key quotes, and answers your questions only from what was actually said. The article starts at the quote, not at the rewind button.

Transcript TXT · SRT

  • 00:38 We have one more announcement before questions.
  • 00:41 The budget goes to parliament on Tuesday, and not a day later.
  • 00:47 Transport gets the largest share for the third year running.
  • 00:55 The details will be published this afternoon.
“The budget goes to parliament on Tuesday, and not a day later.”
Key moment · 00:41

Ask the recording

One striking sentence for each topic, with the time.

  1. 00:41The budget goes to parliament on Tuesday, and not a day later.
  2. 03:12Every new tram line will be tendered in public.
  3. 07:56We will not raise local taxes next year.

Example values from an invented press conference.

Checkable quotes
Every line keeps its timestamp. One click plays the moment, so you hear the exact wording before it goes to print.
Any language spoken
The transcript is written in the language of the recording, which Lumière detects and names.
Read it in yours
Translate the whole transcript into your language, faithfully: names, numbers and amounts stay exactly as spoken. Download it as TXT or SRT.
Ask in yours
Ask in your own language and the answer comes back in it, with the moments it is based on.
Take the selection
Ask for the parts you need, then take them as text, one audio file or the film. Each comes from your latest request.
Radio
Upload a programme as audio: transcript, selection, and a check against your own rules with loudness measured to EBU R128.

Only need to read, search and quote recordings? LumiVision Press, €19 a month.

Coming Speaker names in the transcript.

Compliance report Re-verified · clean
TimeRuleConfidenceDecision
00:04:12 Unsubstantiated claim 92% Hidden
00:17:38 Competitor mention 81% Kept · approved
00:41:05 Missing disclosure 74% Hidden

Loudness · EBU R128

Integrated
−23.1 LUFS
True peak
−1.4 dBTP
Range
7.8 LU
Verdict
Pass
Example values. The structure is the real report, as PDF or DOCX, with the flags also as CSV.

Comply · Compliance Studio

Compliant. And proven.

The compliance notes have a habit of arriving after picture lock. Compliance Studio reads your own guidelines, scans what is said in the transcript for content issues, and keeps a human in the loop for every passage. You decide what stays. What you hide is left out of the edit in Studio and of the master, never erased from your original, and you can bring it back at any time.

  1. Guidelines in PDF, DOCX, TXT, or a photo of the page
  2. The transcript scanned each flag with a confidence score
  3. A human decides hide or keep each passage
  4. Hidden, not erased your original stays untouched; undo any time
  5. Re-verified the compliant master, re-transcribed and re-scanned

Render the compliant master and we check our own work. Proof, not promises.

Teams · Studio & Broadcast

Everyone on the same cut.

A newsroom is not one editor with a deadline. It is a chief editorial, a camera unit, an editor on the desk, and a press office that has to say yes. Lumière seats all of them around the same footage, each with exactly the reach their role deserves.

The shoot

A news team films a politician’s speech across town. The crew packs up; on the drive back, the footage is already uploading from the field. Nobody is waiting for a card to arrive.

The newsroom, live
The chief editorial and the editor read the transcript on their phones while the car is still on the ring road. The speaker’s press team holds a Reviewer seat: they comment on the exact sentence they mean, and when they approve, the click writes the record itself: “Approved segment 1 (0:00–0:16)”, name, time. They never touch the cut.
The parking spot

By the time the car parks, the story is pre-approved. The edit that used to start when the crew walked in is already past its first sign-off.

Four kinds of seat

  • Owner

    Holds the wallet and the keys. Mints leads; ownership itself is never handed over self-serve.

  • Lead

    The chief editorial seat. Manages the working seats and sees every project, never the ownership.

  • Member

    The working seat. Edits, produces, spends lumens ✦ from the one shared wallet.

  • Reviewer

    The external seat. Views, comments, approves. Never edits a frame of the cut, never buys lumens.

Position labels, the way a film credit reads them

  • Chief editor
  • Head of editing
  • Editor
  • Editorial team
  • Camera unit
  • Production management
  • Staff
  • Revision / Press department

Owners mint leads; leads manage the working seats; every seat spends from the one wallet, and only the owner buys lumens. Nothing extra to buy: collaboration arrives with your seats.

The newsroom machinery

  • Project walls

    Members are assigned per project; leads see everything. Ask for a project that isn’t yours and you get the same 404 as one that doesn’t exist.

  • The project wire

    One quiet, chronological feed per story. Comments pin to the exact word of the transcript, and the pins survive a re-cut. @mention a seat; they get a badge.

  • Named approvals

    “Approved segment 1 (0:00–0:16)”: name and time, written by the approve click itself. The editorial record, not a chat scroll.

  • Emergency bulletin

    One org-wide notice when the schedule breaks. Owner and lead only. Each person dismisses their own.

  • The Project Record

    The whole story exported as one self-contained page that prints to PDF, with the comms as CSV or JSON. Export it any time before a project is deleted.

In development
  • Email digests

    The wire reaching your inbox when you’re away. In-app badges ship today; email is being built.

LumiVision Add-on

Your whole archive, understood.

Lumière is where you edit. LumiVision is where you oversee. One library for all the footage your organisation holds, edited in Lumière or not: every video with a summary, chapters and a searchable transcript. Ask a video a question. Keep the moments that matter in collections. Included with Broadcast, an add-on for other plans.

Your library stays on your own server. LumiVision reads it there and keeps only what it needs to show and search it: posters, keyframes, transcripts and a temporary working cache.

LUMIÈRE Vision LibraryCollections Search every word in the archive…
  • 1:42:10 Press conference · day 2 Summary · chapters · transcript
  • 3:05:44 Festival recap rushes Summary · chapters · transcript
  • 58:31 Founder interview Summary · chapters · transcript
  • 1:12:02 Board briefing, Q3 Summary · chapters · transcript
  • 41:19 Training module 4 Summary · chapters · transcript
  • 2:16:47 Evening panel Summary · chapters · transcript
Example library. The layout and features are real; the footage names are illustrative.

The arsenal

570+ AI models, one gateway.

We choose the right one for each task, and switch the moment a better one arrives. Your lumens never notice. Lumière never trains a model on your content.

  • GPT Image 1 MiniOpenAI
  • AnimateDiffOpen model
  • Happy HorseAlibaba
  • H3 Max 16-bit PixelMiniMax
  • Codestral 2508Mistral
  • Recraft 20bRecraft
  • ControlNet SDXLStability AI
  • BagelByteDance
  • Cosmos 3 SuperNVIDIA
  • PixVerse C1 Text To VideoPixVerse
  • Hunyuan A13B InstructTencent
  • Krea 2 LargeKrea
  • Seed 1.6ByteDance Seed
  • FiboBria
  • SonarPerplexity
  • Ernie ImageBaidu
  • Hidream I1 DevHiDream
  • MiMo-V2.5Xiaomi
  • Ling 3.0 FlashinclusionAI
  • MAI Image 2.5 Pro (Text to Image)Microsoft
  • Step 3.5 FlashStepFun
  • Eleven Multilingual v2ElevenLabs
  • Schematron V2 SmallInference.net
  • Moondream 3 PreviewMoondream
  • Perceptron Mk1Perceptron
  • Relace Apply 3Relace
  • ExtractScaleDown
  • InklingThinking Machines
  • AvatarsVeed
  • Avatars Text to VideoArgil
  • Command A+Cohere
  • GlimpseElorian
  • Ember-1Fireworks
  • LFM2.5-24B-A2BLiquidAI
  • Avatar XMirage Api
  • Playground v2.5Playground
  • MAGI-1 (Distilled)Sand AI
  • Pareto 26.10Unbiased
  • Palmyra X5Writer
  • GPT Image 2 APIOpenAI
  • AnimateDiff TurboOpen model
  • Happy Horse 1.1 Text to VideoAlibaba
  • H3 Max Hand DrawnMiniMax
  • Devstral 2 2512Mistral
  • Recraft V3Recraft
  • FooocusStability AI
  • Bytedance Seedance V1 Pro Fast Text To VideoByteDance
  • Cosmos Predict 2.5 2BNVIDIA
  • PixVerse V3.5 Text To VideoPixVerse
  • Hunyuan ImageTencent
  • Krea 2 MediumKrea
  • Seed 1.6 FlashByteDance Seed
  • Fibo 2.0 GenerateBria
  • Sonar Deep ResearchPerplexity
  • Ernie Image LoraBaidu
  • Hidream I1 FastHiDream
  • MiMo-V2.5-ProXiaomi
  • Ling 3.0 Flash VLinclusionAI
  • Mai Image 2.5 Text to ImageMicrosoft
  • Step 3.7 FlashStepFun
  • Eleven Sound EffectsElevenLabs
  • Schematron V2 TurboInference.net
  • Moondream 3.1Moondream
  • Perceptron Mk1.5Perceptron
  • Relace SearchRelace
  • SummarizeScaleDown
  • Inkling SmallThinking Machines
  • Fabric 1.0Veed
  • Flux 2 FlexBlack Forest Labs
  • Gemini 3 Flash PreviewGoogle
  • Qwen3 14BQwen
  • Kling 1.6Kuaishou
  • GLM 4.5Z.ai
  • DeepSeek V3 0324DeepSeek
  • Claude Opus 4.1Anthropic
  • Ideogram TransparentIdeogram
  • Llama 3.1 8B InstructMeta
  • Ltx 2.5 Text to Video FastLightricks
  • Kimi K2 ThinkingMoonshotAI
  • Grok 4.3SpaceXAI
  • Nova Micro 1.0Amazon
  • Luma Ray 3.2 Text to VideoLuma AI
  • Grok Imagine VideoxAI
  • Heygen Video AgentHeyGen
  • Kandinsky5Sber AI
  • Mercury 3Inception
  • LongCat Video DistilledMeituan
  • Hermes 4 405BNous
  • UnslopNemo 12BTheDrummer
  • FLUX 2 LoraBlack Forest Labs
  • Gemini 3 Pro Image PreviewGoogle
  • Qwen3 235B A22B Instruct 2507Qwen
  • Kling 4.0 Text to VideoKuaishou
  • GLM 4.5 AirZ.ai
  • DeepSeek V3.1DeepSeek
  • Claude Opus 4.5Anthropic
  • Ideogram V2Ideogram
  • Llama 3.2 1B InstructMeta
  • LTX Video (preview)Lightricks
  • Kimi K2.5MoonshotAI
  • Grok 4.5SpaceXAI
  • Nova Premier 1.0Amazon
  • Luma Uni-1 Text to ImageLuma AI
  • Grok Imagine Video 1.5 Lite Text to VideoxAI
  • HeyGen Video Text to VideoHeyGen
  • Kandinsky5 ProSber AI
  • GPT-3.5 Turbo (older v0613)OpenAI
  • Boogu ImageOpen model
  • Qwen ImageAlibaba
  • H3 Max Turbo Text to VideoMiniMax
  • Ministral 3 8B 2512Mistral
  • Recraft V4 ProRecraft
  • Stable Audio 2Stability AI
  • Bytedance Seedream V4.5 Text To ImageByteDance
  • Nemotron 3 SuperNVIDIA
  • PixVerse V4 Text To Video FastPixVerse
  • Hunyuan Video V1.5Tencent
  • Krea 2 Text to Image Turbo StyleKrea
  • Seed-2.0-LiteByteDance Seed
  • Fibo LiteBria
  • Sonar Reasoning ProPerplexity
  • GPT-3.5 Turbo 16kOpenAI
  • Bria Text-to-Image BaseOpen model
  • Qwen Image 2512Alibaba
  • H3 Max VHSMiniMax
  • Mistral Large 3 2512Mistral
  • Recraft V4 Pro (Vector)Recraft
  • Stable CascadeStability AI
  • Seedance 1.0 ProByteDance
  • Nemotron 3 UltraNVIDIA
  • PixVerse V4.5 Text To VideoPixVerse
  • Hy-MT2-1.8BTencent
  • Krea 2 TurboKrea
  • Seed-2.0-MiniByteDance Seed
  • FLUX 2 Lora Gallery RealismBlack Forest Labs
  • Gemini 3.1 Pro PreviewGoogle
  • Qwen3 30B A3B Instruct 2507Qwen
  • Kling LipSync Text-to-VideoKuaishou
  • GLM 4.6VZ.ai
  • DeepSeek V3.2 ExpDeepSeek
  • Claude Opus 4.8Anthropic
  • Ideogram V2A TurboIdeogram
  • Llama 4 MaverickMeta
  • LTX Video-0.9.7 13B DistilledLightricks
  • Kimi K3MoonshotAI
  • Grok Build 0.1SpaceXAI
  • Flux 2 MaxBlack Forest Labs
  • Gemini 3.5 FlashGoogle
  • Qwen3 32BQwen
  • Kling O3 Text to Video [Pro]Kuaishou
  • GLM 4.7Z.ai
  • DeepSeek V4 Flash 0423DeepSeek
  • Claude Opus 5Anthropic
  • Ideogram V4.0 Text to ImageIdeogram
  • Llama 4 ScoutMeta
  • LTX-2.3 22BLightricks
  • Kimi K3.1 ProMoonshotAI
  • Flux 2 ProBlack Forest Labs
  • Gemini 3.5 Flash LiteGoogle
  • Qwen3 Coder 30B A3B InstructQwen
  • Kling O3 Text to Video [Standard]Kuaishou
  • GLM 4.7 FlashZ.ai
  • DeepSeek V4 Flash 0731DeepSeek
  • Claude Opus 5.5Anthropic
  • Ideogram V4.0q Text to Image (LoRA)Ideogram
  • Llama Guard 4 12BMeta
  • LTX-2.3 22B DistilledLightricks
  • GPT-4.1 NanoOpenAI
  • FooocusOpen model
  • Qwen2.5 Coder 32B InstructAlibaba
  • MiniMax (Hailuo AI) Video 01 LiveMiniMax
  • Mistral Medium 3.5Mistral
  • Recraft V4 Styles Text to VectorRecraft
  • Stable Diffusion V3Stability AI
  • Seedance 2.5 US Text to VideoByteDance
  • SanaNVIDIA
  • PixVerse V5.6 Text To VideoPixVerse
  • Hy4 previewTencent
  • Flux 3 ImageBlack Forest Labs
  • Gemini 3.7 FlashGoogle
  • Qwen3 Coder NextQwen
  • Kling VideoKuaishou
  • GLM 5 TurboZ.ai
  • DeepSeek V4 Pro 0423DeepSeek
  • Claude Sonnet 4.6Anthropic
  • Ideogram VectorIdeogram
  • Muse Glimmer 30BMeta
  • GPT-4o (2024-05-13)OpenAI
  • Fooocus InpaintingOpen model
  • Wan 2.5 Text to ImageAlibaba
  • MiniMax H3 Text to VideoMiniMax
  • Mistral Small 3Mistral
  • Recraft V4.1 Text to ImageRecraft
  • Stable Diffusion XL LightningStability AI
  • Seedream 5.0 Pro Text to ImageByteDance
  • Sana v1.5 1.6BNVIDIA
  • GPT-4o (2024-08-06)OpenAI
  • Fooocus Upscale or VaryOpen model
  • Wan 2.5 Text to VideoAlibaba
  • MiniMax H3 Text to Video LoRAMiniMax
  • Mistral Small 3.1 24BMistral
  • Recraft V4.1 Text to Image ProRecraft
  • Stable Video DiffusionStability AI
  • UI-TARS 7BByteDance
  • Sana v1.5 4.8BNVIDIA
  • Flux 3 Text To Video DraftBlack Forest Labs
  • Gemini Omni FlashGoogle
  • Qwen3 VL 235B A22B InstructQwen
  • Kling Video v3 Text to Video [Pro]Kuaishou
  • GLM 5.3Z.ai
  • DeepSeek V4.1 Flash 1017DeepSeek
  • GPT-5OpenAI
  • InfinitalkOpen model
  • Wan Text to VideoAlibaba
  • MiniMax Hailuo 02 [Standard] (Text to Video)MiniMax
  • Mistral Small 4Mistral
  • Recraft V4.1 Text to VectorRecraft
  • GPT-5 MiniOpenAI
  • Infinity StarOpen model
  • Wan v2.2 5BAlibaba
  • MiniMax Hailuo 2.3 [Pro] (Text to Video)MiniMax
  • Mixtral 8x22B InstructMistral
  • Recraft V4.1 Text to Vector ProRecraft
  • Flux LoraBlack Forest Labs
  • Gemma 3 12BGoogle
  • Qwen3 VL 8B InstructQwen
  • Kling Video V3 Turbo Pro Text to VideoKuaishou
  • GLM 5V TurboZ.ai
  • GPT-5.1OpenAI
  • Lumina Image 2Open model
  • Wan v2.6 Text to ImageAlibaba
  • MiniMax M1MiniMax
  • Voxtral Small 24B 2507Mistral
  • GPT-5.1-CodexOpenAI
  • MOSS ImageOpen model
  • Wan v2.6 Text to VideoAlibaba
  • MiniMax M2MiniMax
  • FLUX.1 [dev] Control LoRA DepthBlack Forest Labs
  • Gemma 4 26B A4BGoogle
  • Qwen3.5-27BQwen
  • GPT-5.1-Codex-MiniOpenAI
  • OmniGen v1Open model
  • Wan-2.1 Text-to-VideoAlibaba
  • MiniMax M2.5MiniMax
  • FLUX.1 [dev] with Controlnets and LorasBlack Forest Labs
  • Nano BananaGoogle
  • Qwen3.5-9BQwen
  • GPT-5.2 ChatOpenAI
  • Ovi Text to VideoOpen model
  • Wan-2.2 Text-to-Video A14BAlibaba
  • MiniMax M3MiniMax
  • FLUX.1 [schnell]Black Forest Labs
  • Nano Banana 2Google
  • Qwen3.6 35B A3BQwen
  • GPT-5.3-CodexOpenAI
  • Phota Text to ImageOpen model
  • Z Image BaseAlibaba
  • GPT-5.4OpenAI
  • PixArt-ΣOpen model
  • Z Image Base LoraAlibaba
  • GPT-5.4 MiniOpenAI
  • Pony V7Open model
  • Z Image TurboAlibaba
  • FLUX.1 Krea [dev] with LoRAsBlack Forest Labs
  • Nano Banana 2.1Google
  • GPT-5.4 ProOpenAI
  • Sensenova U1 InfographicOpen model
  • GPT-5.5OpenAI
  • Stable Diffusion with LoRAsOpen model
  • GPT-5.6 LunaOpenAI
  • T2V Turbo - Video CrafterOpen model
  • GPT-5.6 Luna ProOpenAI
  • VecglypherOpen model
  • FLUX.2 [klein] 4B Base LoRABlack Forest Labs
  • FLUX.2 [klein] 4B LoRABlack Forest Labs
  • FLUX.2 [klein] 9BBlack Forest Labs
  • FLUX.2 [klein] 9B BaseBlack Forest Labs
  • FLUX.2 [klein] 9B Base LoRABlack Forest Labs
  • FLUX.2 [klein] 9B LoRABlack Forest Labs
  • FLUX.3 KleinBlack Forest Labs
  • FLUX1.1 [pro]Black Forest Labs
  • FLUX1.1 [pro] ultraBlack Forest Labs
  • FLUX1.1 [pro] ultra Fine-tunedBlack Forest Labs
  • Juggernaut Flux BaseBlack Forest Labs
  • Juggernaut Flux Base LoRABlack Forest Labs
  • Juggernaut Flux LightningBlack Forest Labs
  • Juggernaut Flux ProBlack Forest Labs
  • Rundiffusion Photo FluxBlack Forest Labs
  • gpt-oss-safeguard-20bOpenAI
  • Whisper large-v3 turboOpenAI
  • FLUX 2Black Forest Labs
  • Gemini 2.5 Flash ImageGoogle
  • Qwen2.5 7B InstructQwen
  • Kling 1.0Kuaishou
  • CogVideoX-5BZ.ai
  • DeepSeek Janus-ProDeepSeek
  • Claude Haiku 4.5Anthropic
  • IdeogramIdeogram
  • DemucsMeta
  • Ltx 2.3 QualityLightricks
  • Kimi K2 0711MoonshotAI
  • Grok 4.20SpaceXAI
  • Nova 2 LiteAmazon
  • Luma PhotonLuma AI
  • Grok Imagine ImagexAI
  • HeygenHeyGen
  • Kandinsky 6.0 LiteSber AI
  • Mercury 2Inception
  • Longcat ImageMeituan
  • Hermes 3 405B InstructNous
  • Cydonia 24B V4.1TheDrummer
  • Granite 4.0 MicroIBM
  • Mistral LargeMistral AI
  • Morph V3 FastMorph
  • Reka EdgeRekaai
  • Llama 3 8B LunarisSao10K
  • ViduShengshu
  • Solar Mini 4Upstage
  • Magnum v4 72BAnthracite Org
  • Emu 3.5 ImageBAAI
  • BorealCreatify
  • AuraFlowfal
  • MythoMax 13BGryphe
  • Weaver (alpha)Mancer
  • Marey Realism V1.5Moonvalley
  • Beam 1.0 501B A23BReflection
  • Tesseract OCRTesseract
  • ReMM SLERP 13BUndi95
  • Xing4.0 29B A4BXingChen
  • FLUX 2 FlashBlack Forest Labs
  • Gemini 2.5 Pro Preview 06-05Google
  • Qwen2.5 VL 72B InstructQwen
  • Kling 1.5Kuaishou
  • CogViewZ.ai
  • DeepSeek V3DeepSeek
  • Claude Haiku 5.5Anthropic
  • Ideogram Text to ImageIdeogram
  • Llama 3.1 70B InstructMeta
  • LTX 2.3 Video FastLightricks
  • Kimi K2 0905MoonshotAI
  • Grok 4.20 Multi-AgentSpaceXAI
  • Nova Lite 1.0Amazon
  • Luma Photon FlashLuma AI
  • Grok Imagine Image 2.0xAI
  • Heygen v5 Digital TwinHeyGen
  • Kandinsky 6.0 ProSber AI
  • Mercury 2.5Inception
  • LongCat VideoMeituan
  • Hermes 3 70B InstructNous
  • Skyfall 36B V2TheDrummer
  • Granite 4.2 8BIBM
  • Mistral Large 2407Mistral AI
  • Morph V3 LargeMorph
  • Reka Flash 3Rekaai
  • Llama 3.1 Euryale 70B v2.2Sao10K
  • Vidu Text to VideoShengshu
  • Solar Pro 4Upstage
  • GPT Image 2.5 Flare Text to ImageOpenAI
  • Bernini-R Text to VideoOpen model
  • Krea Wan 14b- Text to VideoAlibaba
  • H3 Max Low PolyMiniMax
  • Ministral 3 14B 2512Mistral
  • Recraft V4Recraft
  • SDXL ControlNet UnionStability AI
  • Bytedance Seedance V1.5 Pro Text To VideoByteDance
  • Cosmos Predict 2.5 2B DistilledNVIDIA
  • PixVerse V3.5 Text To Video FastPixVerse
  • Hunyuan Image 3.0 InstructTencent
  • Krea 2 Medium Text to Image TurboKrea
  • Seed 2.1 TurboByteDance Seed
  • Fibo Bbq PreviewBria
  • Sonar ProPerplexity
  • Ernie Image Lora TurboBaidu
  • Hidream I1 FullHiDream
  • MiMo-V2.6-FlashXiaomi
  • Ling 3.1 FlashinclusionAI
  • Phi 4Microsoft
  • Step 5 PreviewStepFun
  • GPT Image 2.5 Sunburst Text to ImageOpenAI
  • BitdanceOpen model
  • Ovis ImageAlibaba
  • H3 Max Retro Toon 70sMiniMax
  • Ministral 3 3B 2512Mistral
  • Recraft V4 (Vector)Recraft
  • SoteDiffusionStability AI
  • Bytedance Seedream V4 Text To ImageByteDance
  • Nemotron 3 Nano 30B A3BNVIDIA
  • PixVerse V4 Text To VideoPixVerse
  • Hunyuan VideoTencent
  • Krea 2 Text to Image Turbo LoRAKrea
  • Seed-2.0-CodeByteDance Seed
  • Fibo Gen 1.5 Text to ImageBria
  • Sonar Pro SearchPerplexity
  • Ernie Image TurboBaidu
  • Hidream O1 ImageHiDream
  • MiMo-V2.6-ProXiaomi
  • Flux 2 Lora GalleryBlack Forest Labs
  • Gemini 3.1 Flash Image PreviewGoogle
  • Qwen3 235B A22B Thinking 2507Qwen
  • Kling ImageKuaishou
  • GLM 4.5VZ.ai
  • DeepSeek V3.1 TerminusDeepSeek
  • Claude Opus 4.6Anthropic
  • Ideogram V2 TurboIdeogram
  • Llama 3.2 3B InstructMeta
  • LTX Video 2.3 ProLightricks
  • Kimi K2.6MoonshotAI
  • Grok 4.6SpaceXAI
  • Nova Pro 1.0Amazon
  • Luma Uni-1 Text to Image MaxLuma AI
  • Grok Imagine Video 1.5 Text to VideoxAI
  • FLUX 2 Lora Gallery Hdr StyleBlack Forest Labs
  • Gemini 3.1 Flash LiteGoogle
  • Qwen3 30B A3BQwen
  • Kling LipSync Audio-to-VideoKuaishou
  • GLM 4.6Z.ai
  • DeepSeek V3.2DeepSeek
  • Claude Opus 4.7Anthropic
  • Ideogram V2AIdeogram
  • Llama 3.3 70B InstructMeta
  • LTX Video-0.9.5Lightricks
  • Kimi K2.7 CodeMoonshotAI
  • Grok 4.7SpaceXAI
  • GPT-4OpenAI
  • Bria Text-to-Image FastOpen model
  • Qwen Image 3 Text to ImageAlibaba
  • MiniMax (Hailuo AI) Text to ImageMiniMax
  • Mistral Large 4Mistral
  • Recraft V4 Styles Pro Text to ImageRecraft
  • Stable Diffusion 3.5 LargeStability AI
  • Seedance 2.0 Fast Text to VideoByteDance
  • Nemotron 3.5 Content SafetyNVIDIA
  • PixVerse V4.5 Text To Video FastPixVerse
  • Hy-MT2-30B-A3BTencent
  • Krea 3 Text to ImageKrea
  • GPT-4.1OpenAI
  • Bria Text-to-Image HDOpen model
  • Qwen Image MaxAlibaba
  • MiniMax (Hailuo AI) Video 01MiniMax
  • Mistral Medium 3Mistral
  • Recraft V4 Styles Pro Text to VectorRecraft
  • Stable Diffusion 3.5 MediumStability AI
  • Seedance 2.0 Mini Text to VideoByteDance
  • Nemotron 3.5 LightningNVIDIA
  • PixVerse V5 Text To VideoPixVerse
  • Hy-MT2-7BTencent
  • GPT-4.1 MiniOpenAI
  • DreamshaperOpen model
  • Qwen2.5 72B InstructAlibaba
  • MiniMax (Hailuo AI) Video 01 DirectorMiniMax
  • Mistral Medium 3.1Mistral
  • Recraft V4 Styles Text to ImageRecraft
  • Stable Diffusion v1.5Stability AI
  • Seedance 2.0 US Text to VideoByteDance
  • Nemotron 3.5 Super VLNVIDIA
  • PixVerse V5.5 Text To VideoPixVerse
  • Hy3Tencent
  • FLUX 2 TurboBlack Forest Labs
  • Gemini 3.6 FlashGoogle
  • Qwen3 Coder 480B A35BQwen
  • Kling v2.5 Text to VideoKuaishou
  • GLM 5Z.ai
  • DeepSeek V4 Flash Vision ExpDeepSeek
  • Claude Sonnet 4.5Anthropic
  • Ideogram V4.5 Text to ImageIdeogram
  • Meta Muse Image Text to ImageMeta
  • LTX-Video 13B 0.9.8 DistilledLightricks
  • GPT-4oOpenAI
  • Fooocus Image PromptOpen model
  • WanAlibaba
  • MiniMax H3 Max Text to VideoMiniMax
  • Mistral NemoMistral
  • Recraft V4.1 Flash Text to ImageRecraft
  • Stable Diffusion XLStability AI
  • SeedreamByteDance
  • Sana SprintNVIDIA
  • PixVerse V6 Text To VideoPixVerse
  • FLUX 3 Max Text to VideoBlack Forest Labs
  • Gemini 3.8 FlashGoogle
  • Qwen3 Next 80B A3B InstructQwen
  • Kling Video v2.6 Text to VideoKuaishou
  • GLM 5.1Z.ai
  • DeepSeek V4 Pro 0813DeepSeek
  • Claude Sonnet 5Anthropic
  • V4.0q [fast]Ideogram
  • Muse Spark OpenMeta
  • Flux 3 Text to VideoBlack Forest Labs
  • Gemini 4 ArgonGoogle
  • Qwen3 Next 80B A3B ThinkingQwen
  • Kling Video V3 Standard Turbo Text to VideoKuaishou
  • GLM 5.2Z.ai
  • DeepSeek V4.1 FlashDeepSeek
  • Claude Sonnet 5.5Anthropic
  • V4.0q [instant]Ideogram
  • GPT-4o-miniOpenAI
  • Illusion DiffusionOpen model
  • Wan 3.0 PrimeAlibaba
  • MiniMax Hailuo 02 [Pro] (Text to Video)MiniMax
  • Mistral Small 3.2 24BMistral
  • Recraft V4.1 Text to Image UtilityRecraft
  • Stable Video Diffusion TurboStability AI
  • Flux Kontext LoraBlack Forest Labs
  • Gemini Omni Flash 1.1 Text to VideoGoogle
  • Qwen3 VL 30B A3B InstructQwen
  • Kling Video v3 Text to Video [Standard]Kuaishou
  • GLM 5.3 FlashZ.ai
  • R1 0528DeepSeek
  • Flux Krea LoraBlack Forest Labs
  • Gemma 2 27BGoogle
  • Qwen3 VL 30B A3B ThinkingQwen
  • Kling Video V3 Text to Video 4KKuaishou
  • GLM 5.3 FlashXZ.ai
  • GPT-5 NanoOpenAI
  • Latent Consistency Models (v1.5/XL)Open model
  • Wan v2.2 A14B Text-to-Image A14B with LoRAsAlibaba
  • MiniMax Hailuo 2.3 [Standard] (Text to Video)MiniMax
  • SabaMistral
  • Recraft V4.1 Utility Text to ImageRecraft
  • FLUX.1 [dev]Black Forest Labs
  • Gemma 3 27BGoogle
  • Qwen3.5 397B A17BQwen
  • KolorsKuaishou
  • GLM ImageZ.ai
  • FLUX.1 [dev] Control LoRA CannyBlack Forest Labs
  • Gemma 3 4BGoogle
  • Qwen3.5-122B-A10BQwen
  • GPT-5.1-Codex-MaxOpenAI
  • Nucleus ImageOpen model
  • Wan-2.1 Pro Text-to-VideoAlibaba
  • MiniMax M2.1MiniMax
  • FLUX.1 [dev] Inpainting with LoRAsBlack Forest Labs
  • Gemma 4 31BGoogle
  • Qwen3.5-35B-A3BQwen
  • GPT-5.2OpenAI
  • Omnigen V2Open model
  • Wan-2.1 Text-to-Video with LoRAsAlibaba
  • MiniMax M2.7MiniMax
  • FLUX.1 [dev] with LoRAsBlack Forest Labs
  • Nano Banana (Gemini 2.5 Flash Image)Google
  • Qwen3.6 27BQwen
  • GPT-5.2-CodexOpenAI
  • PATINAOpen model
  • Wan-2.2 Text-to-Video A14B with LoRAsAlibaba
  • MiniMax M3.1 FlashMiniMax
  • FLUX.1 Kontext [max]Black Forest Labs
  • Nano Banana 2 (Gemini 3.1 Flash Image)Google
  • Qwen3.8 2.4T A95BQwen
  • FLUX.1 Kontext [pro]Black Forest Labs
  • Nano Banana 2 LiteGoogle
  • Qwen3.8 27BQwen
  • FLUX.1 Krea [dev]Black Forest Labs
  • Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)Google
  • GPT-5.4 NanoOpenAI
  • Realistic VisionOpen model
  • Z Image Turbo LoraAlibaba
  • FLUX.1 SRPO [dev]Black Forest Labs
  • Nano Banana LiteGoogle
  • FLUX.1 SubjectBlack Forest Labs
  • Nano Banana ProGoogle
  • FLUX.2 [klein] 4BBlack Forest Labs
  • Nano Banana Pro (Gemini 3 Pro Image)Google
  • FLUX.2 [klein] 4B BaseBlack Forest Labs
  • GPT-5.6 SolOpenAI
  • GPT-5.6 Sol ProOpenAI
  • GPT-5.6 TerraOpenAI
  • GPT-5.6 Terra ProOpenAI
  • GPT-6 AstraOpenAI
  • GPT-6 Astra ProOpenAI
  • GPT-6 LunaOpenAI
  • GPT-6 Luna ProOpenAI
  • GPT-6 SolOpenAI
  • GPT-6 Sol ProOpenAI
  • GPT-6.1 SolOpenAI
  • GPT-6.1 Sol ProOpenAI
  • GPT-Image 1.5OpenAI
  • gpt-image-1OpenAI
  • gpt-oss-120bOpenAI
  • gpt-oss-20bOpenAI
  • Whisper large-v3OpenAI

Model names belong to their makers. Language and vision models appear only where a provider runs them with zero data retention. The list updates itself as models arrive and retire.

For professionals

The heavy lifting, done.
The finish, yours.

We won’t pretend to replace your NLE. You finish where you already work. Lumière gives you a fast timeline and an AI workhorse that builds the first assembly from a single prompt, then hands the cut over as a real project. Do the grind here. Finish it your way.

Hand the cut off to

  • DaVinci Resolve
  • EDL for any editor
  • Premiere Pro Coming
  • Final Cut Pro Coming
Start free

Privacy is the brand

Your footage. Your rules.

Two ways to work. Both keep your footage yours.

Mode one · Upload to our EU cloud

Work from anywhere, even a phone.

  • EU data residency Stored and hosted in the European Union. AI steps send only what each model needs, such as audio for transcription or text, to the AI providers we list as sub-processors.
  • Encrypted in transit Every connection to Lumière runs over TLS.
  • Your original, untouched Lumière never changes, moves or deletes your source file on its own. Compliance hides passages instead of cutting them, and removes its temporary working files after each job.
  • Kept until you say Nothing is deleted unless you ask. Delete a Studio project or a compliance job yourself, or close your account: its content is erased 30 days later.
  • Not our training data Lumière never trains models on your content. Business customers get a GDPR Data Processing Agreement.

Mode two · Lumière Local Early access

Or never upload at all.

Some footage is not allowed to travel: a client contract, a house rule, or your own. With Lumière Local the AI reads only the transcript; your computer does the cutting. The footage never leaves your drive. If speech recognition can't run on your computer, Lumière tells you and sends compressed audio for transcription instead, never the video.

A day of 4K rushes50 GB · stays on your drive
What travels to the cloud~0.4 MB of text

Best for big files and strict-privacy work. Runs on your desktop: the Lumière app for macOS, or the Local Agent for Mac, Windows and Linux. On a phone or tablet, you upload instead.

Get early access

For everyone

One pipeline. Every storyteller.

  • Broadcast stations Compliant masters with verified output, and a content-compliance audit trail.
  • Radio stations Programmes as audio: transcripts, the selection as one sound file, and checks against your own rules with loudness to EBU R128.
  • News agencies Every word of hours of rushes, searchable, and a chat that tells you where the speaker covers your topic.
  • Newspapers & magazines The press conference or interview as a transcript with timestamps, a summary with its key quotes, and answers from what was said.
  • Production companies A junior editor that drafts the cut overnight.
  • Film & documentary Documentary story arcs from hundreds of minutes of rushes.
  • Vloggers & creators Upload, say what you’re making, get a first cut and a thumbnail.
  • Event companies Same-day event recap films.
  • Corporate & training Training videos with your logo and colours, subtitled in more than 30 languages.
  • Personal Your vacation footage, edited like a film. No editing skills required.

How we compare

We combine what others split apart.

Good tools already exist for captions, shorts and search. Where they lead, we say so. Lumière’s difference is the whole run: raw footage to a first cut, checked against your own guidelines, private by design. That run is the difference.

Capability VEED.ioOpusClipWayin.aiDescriptLumière
Auto captions & subtitles yes yes yes yes yes
Translate to other languages yes no no partial yes
Dub into other languages with stock voices yes no partial partial yes
Word-by-word / animated captions yes yes partial partial yes
Search footage by what was said no partial yes yes yes
Edit by editing the text: delete the words, the video follows partial no yes yes yes
Understands the footage: summary, key moments, chat about it partial no yes partial yes
Auto social shorts / vertical reframe yes yes yes partial manual version shipped; automatic on the roadmap
Story dramaturgy from a written brief, not clip-picking no no no partial yes
Full edit: raw footage → finished first cut partial no no partial yes
Sync an external audio recorder to the video by its sound no no no no yes
Multicam sync by timecode or sound, aligned tracks ride the export no no no no yes
Export a real timeline to your NLE (DaVinci Resolve, or any editor via EDL) no no no yes yes
Loudness QC against EBU R128, measured, with the figures in your report no no no no yes
Runs on-device: footage never leaves your machine (Lumière Local, early access) no no no no in development
Team seats with roles & external reviewer seats partial no no yes yes
Newsroom review: anchored notes & named approvals partial no no partial yes
360° footage: pick the virtual camera in the edit no no no no in development
Remembers named people, animals and objects across footage no no no no in development

✓ yes ~ partial – no soon on our roadmap in dev being built now ✓ / soon manual shipped · automatic on our roadmap

Compared honestly against each product’s public feature pages. Capabilities shift, and where a rival is strong we’ve said so. We don’t claim what we haven’t shipped.

Pricing

Simple. In lumens.

  • Trial

    €0

    • 15 lumens ✦, once, after you confirm your email
    • Studio and Compliance
    • No card needed
    Start free
  • Personal

    €19 /mo

    • 60 lumens ✦ per month
    • Prepaid lumen packs
    • Live lumen balance and a price before every job
    Start free
  • Creator

    €49 /mo

    • 200 lumens ✦ per month
    • AI image and music generation, priced in lumens
    • Prepaid lumen packs
    Start free
  • Broadcast / Enterprise

    €499 /mo

    starting price, annual contract

    • 3,500 lumens ✦ contracted
    • Seats at scale: the whole team, one wallet
    • LumiVision included
    • SLA and DPA in the contract
    • Onboarding included
    Talk to us

LumiVision Press For journalists: €19/mo with 120 lumens ✦, transcripts, subtitles in your language and answers with the exact quote. LumiVision only. See LumiVision Press

What is a lumen?

Lumière runs on light, so the currency is the lumen, the unit projectors are measured in. One balance for everything: a minute of footage through the pipeline ≈ 1 ✦, a thumbnail, a dub: each shows its exact price before you click. Your balance is always current; a job that needs more than you have stops and asks you to top up first, never a surprise bill. You spend light to make films.

Trust

Questions, answered.

Where is my footage stored?

In the European Union, and every connection is encrypted in transit. AI steps send only what each model needs, such as audio for transcription or text, to the AI providers we list as sub-processors. Lumière never changes, moves or deletes your original file on its own. Compliance jobs remove their temporary working files after each job. A Studio project or a compliance job stays until you delete it, and the account owner can close the account in Settings.

Do I have to upload my footage?

It depends how you work. From a phone or tablet, you upload in resumable chunks, removed once your file is put back together. On a computer, Lumière Local (early access) removes the upload entirely: processing runs on your own machine and only the transcript, kilobytes instead of gigabytes, reaches the cloud, so your footage never leaves your drive. If speech recognition can’t run on your machine, Lumière tells you and sends compressed audio for transcription instead, never the video. Lumière Local runs on Mac, Windows and Linux; the Lumière desktop app is for macOS.

Can I use Lumière only for transcripts and quotes?

Yes. Upload a recorded video or audio file and Lumière writes the transcript with timestamps, which you can search, translate into your language and download as TXT or SRT. A summary gives the chapters and key quotes, and Ask answers only from what was said, in the language you ask in. Ask for the parts you need and take them as text, one audio file or the film. Check each quote against the recording before you publish it: one click plays the moment. Speaker names are coming.

Does Compliance cut my original footage?

No. Compliance marks the passages that break your guidelines; you decide which to hide. Hidden passages are left out of what you deliver: the edit you send to Studio, its renders and NLE exports, and the optional compliant master. Your original file is never changed, and you can un-hide a passage at any time, also after the hand-off; the Studio project follows.

What happens when I run out of lumens?

Every job shows its price first. If it needs more lumens than you have, it stops and asks you to top up before it runs, on every plan. Nothing is ever charged beyond your prepaid lumens. Prepaid lumen packs are one click away, and your work continues where it was. On a team, the owner buys the lumens.

What internet speed do I need?

Almost none. With Lumière Local, only kilobytes travel: a transcript up (~0.4 MB per hour of footage), an edit plan down. The heavy AI traffic runs on our servers, never over your connection. If your connection can send a text message, it can run a production: hotel Wi-Fi, a tethered phone, a satellite link. AI decisions do need a connection; rendering and cutting don’t.

What models power Lumière?

A routing gateway picks the right model for each task: fast models for transcription-scale work, premium models for story and edit decisions. Model keys never leave our gateway, and every job shows its lumen price before it runs.

Can I cancel anytime?

Yes. Monthly plans have no minimum term. To cancel, ask us in the help chat or by email; cancelling in your account yourself is coming. Prepaid lumens never expire, also after a plan ends. Broadcast / Enterprise runs on an annual contract.

How does invoicing work?

Every lumen pack purchase automatically generates a VAT invoice. EU VAT is handled at checkout, including the reverse charge for a valid EU business VAT ID.

Can I delete my data or close my account?

Yes. Delete project in Studio and Delete job in Compliance remove a project or a job together with the file you uploaded for it; both ask first and cannot be undone. The account owner can close the account in Settings: everyone is signed out at once, unspent lumens are forfeited, a paid plan ends at the end of its period, and the content is erased 30 days later unless you ask us to reopen it. Invoices and payment records are kept for 8 years, as Hungarian accounting law requires.

Is my content used to train AI models?

Lumière never trains models on your content. The AI providers that process it for you are named in the Privacy Policy, with what each one receives. A GDPR Data Processing Agreement is available on request.