Programmable editing

for agents

Programmable editing

for agents

Turn intent into timeline operations. Compose streams server side, generate missing assets when needed, and ship outputs without rebuilding MP4s.

Turn intent into timeline operations. Compose streams server side, generate missing assets when needed, and ship outputs without rebuilding MP4s.

If an agent can find it, it can edit it.

If an agent can find it, it can edit it.

timeline.compose()

Video

Overlay

Audio

0:00

0:15

0:30

0:45

1:00

intro.…

scene_01.m…

outro scenes

lo…

caption.srt

voiceover.wav

music.mp3

Live Output

Streaming

00:12:34

stream.videodb.io/v1/compose/abc123

timeline.compose()

Video

Overlay

Audio

0:00

0:30

0:45

1:00

intro.…

scene_01.m…

outro scenes

lo…

caption.srt

voiceover.wav

music.mp3

Live Output

Streaming

00:12:34

stream.videodb.io/v1/compose/abc123

timeline.compose()

Video

Overlay

Audio

0:00

0:15

0:30

0:45

1:00

intro.…

scene_01.m…

outro scenes

lo…

caption.srt

voiceover.wav

music.mp3

Live Output

Streaming

00:12:34

stream.videodb.io/v1/compose/abc123

timeline.compose()

Video

Overlay

Audio

0:00

0:15

0:30

0:45

1:00

intro.…

scene_01.m…

outro scenes

lo…

caption.srt

voiceover.wav

music.mp3

Live Output

Streaming

00:12:34

stream.videodb.io/v1/compose/abc123

Code-first timeline
Over 21 Completed
Streams in, Streams out
Over 21 Completed
Non-destructive versions
Over 21 Completed

From prompt to timeline to output

From prompt to timeline to output

From prompt to timeline to output

Editing is not a UI. It is a deterministic graph your agent can generate and rerun.

Editing is not a UI. It is a deterministic graph your agent can generate and rerun.

See

See

Search results, scene windows, metadata

Search result includes a playable window

Rich metadata for every frame

Plan

Plan

Agent decides edits

Agent emits structured edit intent

Deterministic operation graph

Compose

Compose

Timeline operations

Operations apply to a timeline canvas

Non-destructive by default

Deliver

Deliver

Playable stream, export, webhook

Stream variants and exports

Webhook triggers on completion

Operations library your agent can call

Operations library your agent can call

Operations library your agent can call

The agent does not edit frames. It edits a timeline.

The agent does not edit frames. It edits a timeline.

The agent does not edit frames. It edits a timeline.

Trim & cut

Build highlight reels and removals

Stitch & reorder

Compile multi-source stories

Overlay assets

Logos, callouts, UI labels

Blur & redact

Faces, plates, sensitive regions

Captions & subtitles

Generate and place correctly

Trim & cut

Build highlight reels and removals

Stitch & reorder

Compile multi-source stories

Overlay assets

Logos, callouts, UI labels

Blur & redact

Faces, plates, sensitive regions

Captions & subtitles

Generate and place correctly

Resize & reframe

16:9 to 9:16, safe crops

Filters & grading

Consistent look at scale

Transitions

Fades and smooth handoffs

Audio mix

Ducking, mute windows, voice over

Versioning

Keep variants without rewriting base

Generate what's missing, then compose it

Generate what's missing, then compose it

Generate what's missing, then compose it

When footage is incomplete, your agent can generate assets (voice, music, sound effects, images) and drop them into the same timeline.

When footage is incomplete, your agent can generate assets (voice, music, sound effects, images) and drop them into the same timeline.

When footage is incomplete, your agent can generate assets (voice, music, sound effects, images) and drop them into the same timeline.

Generate voice over for a segment

Generate voice over for a segment

Generate music bed and loop to duration

Generate music bed and loop to duration

Generate sound effects for key moments

Generate sound effects for key moments

Generate images for overlays and slates

Generate images for overlays and slates

Video

scene_01.mp4

scene_02.mp4

Audio

✨ Generated VO

music.mp3

SFX

✨ SFX

Ship variants without rebuild pain

Ship variants without rebuild pain

Ship variants without rebuild pain

Real-time compositing means you stream instantly. Export only when you need to.

Real-time compositing means you stream instantly. Export only when you need to.

Traditional Pipeline

0

%

Edit

Render

Re-encode

Upload

Repeat

~15 minutes per variant

VideoDB Pipeline

Ready!

Edit ops

Compose server-side

Stream instantly

Export when needed

Instant streaming available

Traditional Pipeline

0

%

Edit

Render

Re-encode

Upload

Repeat

~15 minutes per variant

VideoDB Pipeline

Ready!

Edit ops

Compose server-side

Stream instantly

Export when needed

Instant streaming available

Traditional Pipeline

0

%

Edit

Render

Re-encode

Upload

Repeat

~15 minutes per variant

VideoDB Pipeline

Ready!

Edit ops

Compose server-side

Stream instantly

Export when needed

Instant streaming available

Blueprints for agentic editing

Blueprints for agentic editing

Blueprints for agentic editing

Reference workflows showing how triggers flow through logic to produce outputs.

Reference workflows showing how triggers flow through logic to produce outputs.

Compliance Redaction

Trigger

Event detects sensitive object

Logic

Blur window + log metadata

Output

Compliant stream variant

Personalized Variants

Trigger

User context received

Logic

Swap overlays, callouts, language

Output

Per-user stream variant

Auto Highlights

Trigger

"High entropy moments" or query

Logic

Stitch best segments, add captions

Output

Shareable reel

Index result

Index result

idx_001

13:15:51

Edit plan

Edit plan

plan_002

13:15:51

Timeline ops

Timeline ops

ops_003

13:15:51

Output URL

Output URL

out_004

13:15:51

Output

stream.videodb.io/v1/compose/abc123

Built for correctness, not demos

Built for correctness,

not demos

Production-grade infrastructure that you can trust and debug.

Production-grade infrastructure that you can trust and debug.

Deterministic timeline operations

Deterministic timeline operations

Debuggable: every output ties back to ops

Debuggable: every output ties back to ops

Versionable: keep variants and promote to canonical

Versionable: keep variants and promote to canonical

Composable: search outputs feed editing directly

Composable: search outputs feed editing directly

Automatable: events can trigger edits

Automatable: events can trigger edits

Give your agent hands,
not suggestions

Give your agent hands,
not suggestions

Give your agent hands,
not suggestions

VideoDB turns understanding into action by making editing a programmable interface.

VideoDB turns understanding into action by making editing a programmable interface.

Python

  from videodb.editor import Timeline, Track, Clip, VideoAsset

  timeline = Timeline(conn)
  timeline.resolution = "1920x1080"

  video_track = Track()

  intro_clip = Clip(
      asset=VideoAsset(id=intro.id, start=0), duration=5)
  main_clip = Clip(
      asset=VideoAsset(id=main.id, start=10), duration=60)
  outro_clip = Clip(
      asset=VideoAsset(id=outro.id, start=0),
      duration=3
  )
video_track.add_clip(start=0, clip=intro_clip)
  video_track.add_clip(start=5, clip=main_clip)
  video_track.add_clip(start=65, clip=outro_clip)

  timeline.add_track(video_track)

  stream_url = timeline.generate_stream()
  from videodb.editor import Timeline, Track, Clip, VideoAsset

  timeline = Timeline(conn)
  timeline.resolution = "1920x1080"

  video_track = Track()

  intro_clip = Clip(
      asset=VideoAsset(id=intro.id, start=0), duration=5)
  main_clip = Clip(
      asset=VideoAsset(id=main.id, start=10), duration=60)
  outro_clip = Clip(
      asset=VideoAsset(id=outro.id, start=0),
      duration=3
  )
video_track.add_clip(start=0, clip=intro_clip)
  video_track.add_clip(start=5, clip=main_clip)
  video_track.add_clip(start=65, clip=outro_clip)

  timeline.add_track(video_track)

  stream_url = timeline.generate_stream()

Python

  from videodb.editor import Timeline, Track, Clip, VideoAsset

  timeline = Timeline(conn)
  timeline.resolution = "1920x1080"

  video_track = Track()

  intro_clip = Clip(
      asset=VideoAsset(id=intro.id, start=0), duration=5)
  main_clip = Clip(
      asset=VideoAsset(id=main.id, start=10), duration=60)
  outro_clip = Clip(
      asset=VideoAsset(id=outro.id, start=0),
      duration=3
  )
video_track.add_clip(start=0, clip=intro_clip)
  video_track.add_clip(start=5, clip=main_clip)
  video_track.add_clip(start=65, clip=outro_clip)

  timeline.add_track(video_track)

  stream_url = timeline.generate_stream()

FAQs

What does “programmable editing” mean in VideoDB?

It means editing is a callable API surface, not a human timeline UI. Your app or agent expresses edits as deterministic timeline operations, and VideoDB produces a playable stream or export from those operations.

Is this meant to replace tools like Adobe?

What is the core unit of editing?

Can an agent go from “find” to “edit” in one flow?

What does “programmable editing” mean in VideoDB?

Is this meant to replace tools like Adobe?

What is the core unit of editing?

Can an agent go from “find” to “edit” in one flow?

What does “programmable editing” mean in VideoDB?

It means editing is a callable API surface, not a human timeline UI. Your app or agent expresses edits as deterministic timeline operations, and VideoDB produces a playable stream or export from those operations.

Is this meant to replace tools like Adobe?

What is the core unit of editing?

Can an agent go from “find” to “edit” in one flow?

What does “programmable editing” mean in VideoDB?

It means editing is a callable API surface, not a human timeline UI. Your app or agent expresses edits as deterministic timeline operations, and VideoDB produces a playable stream or export from those operations.

Is this meant to replace tools like Adobe?

What is the core unit of editing?

Can an agent go from “find” to “edit” in one flow?

The Perception Layer for AI

Apt 2111 Lansing Street San Francisco, CA 94105 USA

HD-239, WeWork Prestige Atlanta, 80 Feet Main Road, Koramangala I Block, Bengaluru, Karnataka, 560034

sales@videodb.com

The Perception Layer for AI

Apt 2111 Lansing Street San Francisco, CA 94105 USA

HD-239, WeWork Prestige Atlanta, 80 Feet Main Road, Koramangala I Block, Bengaluru, Karnataka, 560034

sales@videodb.com

The Perception Layer for AI

Apt 2111 Lansing Street San Francisco, CA 94105 USA

HD-239, WeWork Prestige Atlanta, 80 Feet Main Road, Koramangala I Block, Bengaluru, Karnataka, 560034

sales@videodb.com