One platform.
Three modes of working.
Switch instantly between fast iteration, full-film overview, and the technical graph. The same Shot Groups appear in all three.
Storyboard
Read your film top-to-bottom as a vertical strip of shots. Drag to reorder. Click Play for a 24fps animatic. Apply a Delivery Preset to the whole Sequence in one click.
- Sequences for Acts & Scenes
- Animatic playback at 24fps
- Apply Delivery Preset · whole Sequence
- Comment threads on every Shot Group
Brief it the way
you'd brief a DP.
Up to 9 references per shot. Style refs match the look. Identity refs lock the person — must originate inside the platform (Seedream image) for content-policy compliance.
Text → Video
Brief in, motion out. Starting from a prompt.
Image → Video
Animate a still. Or pin first/last frames as anchors.
Video → Video
Restyle or style-match an existing clip.
Image
Keyframe or concept. Seedream 5.0. No engine tiering.
Extend
Continue a shot +3s or +5s. Anchor on last frame.
Four engines.
Pay for what you actually need.
Iterate cheap on a draft engine. Lock the shot. Render the master on Cinema only when you're ready. The default — Standard, Seedance 2.0 fast at 720p — covers most production work and burns ~½ the VX of a Cinema render.
The default. 2.0 quality at fast speed and lower cost. Where 80% of your work lives. 480p / 720p.
We speak your DP's
vocabulary.
Nine dropdowns. Lens, capture, lighting — binds to a Shot Group and inherits through every take. ARRI Alexa 35. RED Komodo. Sony Venice 2. T-stops, not aperture sliders. Then one-click Delivery Presets apply to a whole Sequence: ARRI LogC4, Sony S-Log3, RED Log3G10, Netflix, Theatrical DCP, Web.
The same face.
Across every shot.
Two tiers of character lock. Real-Person DNA for actors — compliance-grade, QR-authorized. Stylized DNA for concept characters and visual styles — direct upload.
For real actors. Compliance-grade.
Generate a QR code. Actor scans on their phone, verifies their face, agrees to terms. Authorized once, forever. Stored in BytePlus's private portrait library. Every operator on your studio can use it.
For concept characters. No verification.
Upload 3–5 reference images. Different angles, different lighting. The system creates a Persistent Identity in 30 seconds. Bind to any shot.
What we do well.
What we don't fake.
Cinema-grade buyers can spot a marketing claim from a kilometer out. So we publish what works today, and what doesn't. If a DP can spot the source bit depth on a waveform monitor, the marketing copy matches what the waveform shows.
- 5s cinematic clipsstructured around real cinematography — lens, lighting, motion, mood
- Animate a stillgive it a photo, get a moving shot
- Extensions, up to 3 chainedpreserves character, lighting, motion
- Character locks across a filmDNA Library — same person in every shot
- Style matchinglighting · grade · lens character · camera motion from a reference video
- Mastered files for editorialProRes 4444 XQ · LogC4 · DNxHR · EXR sequence
- Custom LUTsyour show LUT applied at render time
- Lip-sync dialogueADR-replace audio in post. Hedra integration on roadmap.
- Long continuous takes >15scharacter drifts past 3 chained extends. Plan for cuts.
- Hands in tight close-upcurrent AI models still struggle. Avoid or shoot practical.
- Specific real locations“my grandmother’s kitchen” won’t work. Iconic locations will.
- Native 10-bit colorcontainer is 10-bit, source is 8-bit dithered. Won’t claim what the waveform won’t show.
- Real-person likeness without consentQR-authorization is required. Not a policy, an API constraint.
- Fast multi-character actionfight scenes with complex blocking get simplified.
A studio your DPs
can actually run.
Two ways to pay.
Pick the one that fits your studio.
Most studios start with Managed — one bill, VX included. Heavy users with their own ByteDance contracts bring their own key. Both unlock the full platform.
| Feature | Free | Studio Managed | Studio BYOK | Enterprise |
|---|---|---|---|---|
| Price | $0 | $49/mo + VX | $99/mo flat | Custom |
| Who pays BytePlus? | VisionX | VisionX | You (BYOK) | You + VisionX fallback |
| Monthly VX | 60 | 500 + PAYG | Your quota | Both |
| Watermark | Yes | — | — | — |
| Queue priority | Low | Medium | High | Reserved |
| Generation Space | ✓ | ✓ | ✓ | ✓ |
| Workflow Canvas | — | ✓ | ✓ | ✓ |
| Cinema Controls | — | ✓ | ✓ | ✓ |
| DNA · Characters | — | 3 | Unlimited | Unlimited |
| DNA · Style Tokens | — | 5 | Unlimited | Unlimited |
| Custom LUT upload | — | ✓ | ✓ | ✓ |
| All delivery presets | — | — | ✓ | ✓ |
| Multi-user / teams | — | — | — | ✓ |
| SLA | — | — | — | 99.9% |
| Notion sync | — | ✓ | ✓ | ✓ |
| API access | — | — | ✓ | ✓ |
One VX is $0.10.
That's it.
Pinned. Always. A 5-second 720p clip on the Standard engine — the production default — is ~10 VX. Pick a cheaper engine to iterate. Switch to Cinema for the final master. Exact cost shows before every generation.
Five people in the room.
One platform for all of them.
Margin protection. Speed to delivery. The pipeline you’d build in 6 months if you had 2 engineers — built.
Cinema Controls in your vocabulary. Output your colorist can grade. Anamorphic that actually looks anamorphic.
Notion sync. Failures DB → SOPs. VX caps per project. Audit trail on every shot, every operator.
Direct ByteDance API. OCIO color science. JSON export. Source code escrow at Enterprise. Not a black box.
Predictable monthly cost. Cancel any time. Transparent cost buildup. ROI typically 2 months at studio scale.
“Calivision is shipping Mahashakti pre-vis on VisionX today. The output lands in our DI without conformance work, and our DPs can finally direct an AI tool in the vocabulary they already use.”