Time to first frame is one of those numbers nobody talks about until it ruins their workflow. You click a clip, you expect to see the first pixel immediately. If you have to wait three seconds, the editing flow breaks. You start clicking around to verify the file actually loaded. You stop trusting the timeline.
On May 5, 2026, we shipped a major upgrade to the in-house Tellers video player. The headline number is 10x faster time to first frame, alongside fixes for buffering at clip boundaries, missing-frame artifacts, and HTML rendering inside the player. This is a short look at what changed and why it matters for AI video editing.
What Time to First Frame Actually Measures
Time to first frame (TTFF) is the latency between requesting a video and seeing its first decoded frame on screen. It is the sum of several smaller stages:
- Network: fetching the bytes the player needs to start
- Container parsing: extracting metadata so decoding can begin
- Decoding: the codec rendering the first I-frame
- Render scheduling: handing the frame to the GPU and painting it
Each stage is small in isolation. Stacked together, they decide whether your editor feels instant or sluggish.
Why It Matters for AI Video Editing
AI video editing tools move clips around constantly. The agent generates a new shot, swaps a clip, or proposes an alternate take. You preview the result. You iterate.
That loop only works if previewing feels instant. Otherwise the friction shifts from “I am editing” to “I am waiting.” A delay of half a second sounds tiny, but it compounds across hundreds of interactions per session. Worse, it changes how you work — you stop exploring options because every preview costs you patience.
A fast player is not a polish item. It is a load-bearing part of the editing experience.
What Changed in the Tellers Player
The May 5 release upgraded the in-house player with several optimisations:
- 10x faster time to first frame through smarter buffering and a leaner decoding path
- Smooth playback at clip boundaries — no more stalls when transitioning between adjacent clips on the timeline
- Fewer decoding artifacts caused by occasional missing frames due to floating point imprecision
- Faster HTML rendering inside the player for overlays, captions, and HTML-based clips
- More reliable audio playback including fixes for desync issues
Most of these improvements are invisible when they work.
Why We Build Our Own Player
Tellers is an AI video creation and editing platform. A single project can mix multiple clips on multiple layers from multiple source videos from different servers in different codex, the source videos can hours but the player will on fetch the relecant chunks. On top of that we also need to handle editable captions, overlays and soon motion graphics.
We had to build a tailor made video player to enable this realtime cloud based video timeline rendering. You can see the result video as soon as your agent made a change, no rendering time, no downloading delays (on decent connections).
For now we use a lot of RAM to enable all of this to work so fast but we’re working on optimisation that will enable the player to even work on mobile.
Owning the player lets us decide:
- How clips are buffered before the playhead reaches them
- How the player behaves when the agent updates a clip mid-session
- How decoded frames are reused across the timeline
- How HTML, generated visuals, and uploaded footage share the same rendering surface
The May 5 release is one step in that ongoing work. There is more coming, next step reduce RAM usage to enable mobile editing.
Try It
If you already use Tellers, the upgrade is live — open any project and the new player is in your hands.
If you are new to Tellers, start a project and notice how quickly the first frame appears when you click around the timeline. The full platform is at tellers.ai.
What is time to first frame?
Time to first frame (TTFF) is the latency between requesting a video and seeing its first decoded frame on screen. It includes network fetch, container parsing, decoding, and render scheduling.
Why does TTFF matter in an AI video editor?
Because you preview clips constantly. Every cut, swap, and agent iteration triggers a new playback request. If TTFF is slow, the editing loop turns into a waiting loop.
What changed in the Tellers video player?
The May 5, 2026 release shipped a 10x improvement in time to first frame, smoother playback at clip boundaries, fewer decoding artifacts, faster HTML rendering, and more reliable audio.
Do I need to update anything to get the new player?
No. The new player is live for everyone. Open a project at app.tellers.ai and the upgrade is already applied.