Commercial video teams
Create seasonal, regional, material or audience variations from an approved master shot.
Independent tool overview
Ray 3.2 Modify at a glance
Ray 3.2 Modify is Luma's current production-focused video-to-video model for transforming existing footage while preserving selected motion, structure, timing and character details.
Visit the official Ray 3.2 Modify site ↗
- Best for
- Production-controlled transformation of existing footage
- Current model
- Ray 3.2 Modify
- Maximum source length
- 20 seconds
- Keyframes
- Up to 64 at exact source-frame indexes
- Output
- Same duration as the source, with HDR and EXR options
- Consumer access
- Luma plans from $30/month or $300/year
Overview
What Ray 3.2 Modify is
Ray3 Modify began as Luma's controlled video-to-video workflow for changing a clip's character, style or environment with reference images and start or end keyframes. Luma's current successor is Ray 3.2 Modify, which expands the idea into a more production-oriented editor.
Ray 3.2 always starts with a source video and returns a transformed clip with the same duration, up to 20 seconds. Creators can guide the result with a prompt, as many as 64 keyframes placed at exact source frames, or both, then independently control how much motion, scene structure, faces, bodies and poses should be preserved.
This is not a general text-to-video model. Its value comes from reusing a real performance, camera move, product angle, edit or shot layout while changing the visual treatment. Teams needing loops or duration changes may still use Ray3.14 Modify; teams needing multi-keyframe control, HDR, EXR or deeper preservation settings should use Ray 3.2.
Use cases
Who Ray 3.2 Modify is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
VFX and look-development artists
Explore stylized treatments, environment changes and temporary composites before committing to a conventional finishing pipeline.
Directors preserving performances
Retain selected motion, body shape, face, pose and timing while changing the character or world around the performance.
Product and fashion creators
Restyle materials, wardrobe, signage, props, weather and environments without rebuilding the original camera move.
Capabilities
Core Ray 3.2 Modify features
Source-video transformation
Reimagines an existing clip while preserving its exact duration and selected visual structure.
Multi-keyframe direction
Accepts as many as 64 edited stills at arbitrary source-frame positions to anchor the look throughout the shot.
Prompt or visual guidance
Works with a text description, keyframes alone, or both for stronger creative control.
Motion adherence
Controls how closely the output follows the timing and movement behavior of the source footage.
Structure adherence
Controls preservation of framing, geometry, object placement and overall scene layout.
Character controls
Separately locks faces, bodies and skeletal pose behavior when identity, build or performance must survive the restyle.
HDR and EXR
Supports HDR video-to-video and EXR export for color grading and professional finishing.
Speed and quality modes
Lets creators choose faster iteration or a higher-quality inference path for final candidates.
Web and API workflows
Runs through Luma's creative application and is also available as Ray 3.2 video editing through the Luma Agents API.
Process
How the Ray 3.2 Modify workflow works
Step 1
Choose a strong source plate
Use clear, stable, high-quality footage with the motion, framing and timing that should remain in the result.
Step 2
Define what must stay fixed
Decide whether motion, structure, face, body, pose, product placement or camera behavior is the non-negotiable part of the shot.
Step 3
Create a light first pass
Start with a concise target-state prompt and moderate adherence before adding complex controls.
Step 4
Anchor drift with keyframes
Export important source frames, restyle them carefully and attach them at their original indexes where the output needs more art direction.
Step 5
Finish outside the generator
Review frame continuity, identity, logos, hands, text and artifacts, then color-grade, composite, edit and clear rights before release.
Cost
Ray 3.2 Modify pricing and free plan
Luma's individual plans are Plus at $30 monthly or $300 yearly, Pro at $90 monthly or $900 yearly, and Ultra at $300 monthly or $3,000 yearly. Ray 3.2 video-to-video in the web app consumes 36 credits/second in Draft, 48 at 540p, 72 at 720p and 144 at 1080p; HDR is 2x and HDR+EXR is 3x. The separate Agents API charges standard 5-second edits from $0.54 to $2.16 by resolution.
Plus
$30/month or $300/year
The entry paid Luma plan with commercial use and access to Luma and third-party models.
- Commercial use
- Guest collaborator editing
- Ray 3.2 usage consumes plan credits
- Top-up credits available on active paid plans
Pro
$90/month or $900/year
For creators who need substantially more Luma Agent usage.
- Everything in Plus
- 4x Luma Agent usage relative to Plus
- Same per-action credit schedule
- Suitable for heavier iteration
Ultra
$300/month or $3,000/year
The highest public individual tier for intensive creative use.
- Everything in Pro
- 15x Luma Agent usage relative to Plus
- Same per-action credit schedule
- Higher monthly budget for production volume
Agents API
$0.54-$2.16 per standard 5-second video edit
Pay-as-you-go Ray 3.2 editing for developers and automated workflows.
- Draft $0.54 per 5 seconds
- 540p $0.72, 720p $1.08 and 1080p $2.16 per 5 seconds
- HDR costs 2x and HDR+EXR costs 3x
- API billing is separate from Luma subscriptions
Team / Enterprise
Contact for pricing
For shared administration, organizational usage and enterprise services.
- Team management and shared projects
- Usage analytics and shared credits
- SSO on business plans
- Enterprise training and custom fine-tuning options
Pricing checked . Check current pricing at the source ↗
Assessment
Ray 3.2 Modify strengths and limitations
Where it stands out
- Transforms real motion and framing instead of rebuilding a shot from a blank prompt
- Up to 64 timeline keyframes provide unusually granular visual direction
- Separate motion, structure, face, body and pose controls expose meaningful production tradeoffs
- Exact source duration simplifies fitting transformed footage back into an edit
- HDR, EXR and quality modes support more professional finishing workflows
- Both interactive and API access are available with transparent usage rates
What to consider
- Ray 3.2 requires source footage and is not the right model for text-to-video, image animation, clip extension or loop creation
- Source clips are limited to 20 seconds
- Identity, hands, logos, text, fine facial cues and object consistency can still drift between frames
- More keyframes and preservation controls improve direction but add setup time and can conflict with the desired transformation
- 1080p, HDR and EXR iterations consume credits quickly, especially across many variants
- Generative output still requires human review, conventional editing and often compositing before professional delivery
- Creators need permission for source footage, likenesses, trademarks and reference images used in a transformation
Compare
Ray 3.2 Modify alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Content Creator
Kling O1
A multimodal video model for creators who want text, image and video editing controls in Kling's ecosystem.
Explore Kling O1 →Content Creator
Runway Gen-4.5
A broader cinematic generation alternative for teams already using Runway's creative production suite.
Explore Runway Gen-4.5 →Content Creator
Veo 3.1
Google's high-end video model for creators prioritizing strong generation quality and native audio workflows over Ray 3.2's source-preserving controls.
Explore Veo 3.1 →Questions
Ray 3.2 Modify FAQs
What happened to Ray3 Modify?
Ray3 Modify remains part of Luma's model history, but Ray 3.2 is the current production-focused successor for standard video-to-video transformation with multi-keyframe, HDR, EXR and deeper preservation controls.
What does Ray 3.2 Modify do?
It takes an existing source clip and creates a visually transformed version with the same duration while letting the creator preserve selected motion, structure and character attributes.
How long can a Ray 3.2 source video be?
Up to 20 seconds. The output keeps the exact source duration rather than using a separate duration setting.
How many keyframes can Ray 3.2 use?
Luma documents support for up to 64 guide keyframes attached to exact indexes in the source timeline.
Should I use Ray 3.2 or Ray3.14 Modify?
Use Ray 3.2 for standard V2V work, multi-keyframe direction, exact source duration, HDR, EXR and deeper character controls. Luma recommends Ray3.14 Modify when a workflow specifically needs to change duration or create a loop.
How much does Ray 3.2 video editing cost?
In the Luma app, standard V2V uses 36 credits/second in Draft, 48 at 540p, 72 at 720p and 144 at 1080p. In the separate pay-as-you-go API, a standard 5-second edit ranges from $0.54 to $2.16 by resolution.
Does Ray 3.2 support HDR and EXR?
Yes. Ray 3.2 supports HDR video-to-video and EXR export, with higher credit and API costs than standard dynamic range.
Can Ray 3.2 preserve a person's identity?
It provides separate face, body and pose controls designed to preserve identity and performance, but results still need frame-by-frame review because fine features can drift.
Bottom line
Our Ray 3.2 Modify verdict
Ray 3.2 is a meaningful upgrade for creators who found Ray3 Modify promising but too limited for art-directed production. Its exact-duration transformation, 64 keyframes and separate preservation controls make it especially useful for look development, campaign variants and stylized VFX. It remains a generative first pass rather than a replacement for editing, compositing and final quality control.
Visit Ray 3.2 Modify website ↗Similar Content Creator Tools
View all Content Creator AI tools →
ChatGPT Images
ChatGPT Images - OpenAI's new AI image generation system with upgraded editing, text rendering, quality, and dedicated Images creation space
Hunyuan Motion 1.0
Hunyuan Motion 1.0 - Tencent's open-source model for 3D character animations from text prompts
Wan 2.6
Wan 2.6 - Alibaba's new native multimodal model for image and video generation
LTX-2 Model Family
LTX-2 - Lightricks' open-source foundation video model for long-form, high-fidelity outputs
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.