Mirelo SFX 1.6 Is Out and enables Audio Consistent video extension

Extending a video clip with AI has usually come with a tradeoff: even when the image continues smoothly, the sound often breaks the illusion.

You could extend the visuals, but the audio would usually need to be regenerated, looped manually, replaced, or hidden under music. For editors, that meant extra work. For viewers, it often meant a subtle but noticeable jump.

Today, our partner Mirelo is releasing Mirelo SFX 1.6, now available on Tellers. One of its most important new capabilities is audio extension, which allows the Tellers agent to extend AI-generated video while preserving audio continuity.

This is a major step toward AI video generation that feels less like a set of disconnected clips and more like a coherent scene.

Tellers as the Agent and Orchestrator

Tellers is not just calling a single model. It acts as the agent and orchestration layer between the user’s intent, video generation models, audio models, timeline editing, and final composition.

When a user asks to extend a shot, they do not need to manually choose which model should handle the image, which model should handle the audio, or how the results should be aligned on the timeline. They only need to express the creative intent.

The Tellers agent then orchestrates the right tools for the task: video extension models for the visual continuation, Mirelo SFX 1.6 for audio continuity, and Tellers’ own timeline system to assemble the result into an editable video sequence.

What Audio-Consistent Extension Means

When you ask Tellers to extend a clip, the agent can use video generation to continue the image. Until now, the audio side was much harder to preserve. The ambience, texture, and sonic identity of the original clip would often need to be rebuilt separately.

With Mirelo SFX 1.6, the original audio is preserved up to the cut point, and the model generates a continuation that follows it naturally.

That means the sound of wind, machinery, footsteps, room tone, rain, crowd noise, or a designed sound effect can continue with the scene instead of abruptly changing when the generated extension begins.

Why This Matters for AI Video Creation

Continuity is one of the main things that makes a scene believable.

Viewers may forgive a small visual imperfection, but the ear is much less forgiving. A rough audio cut, a mismatched ambience, or a repeated loop can instantly make a generated sequence feel artificial.

Audio-consistent extension unlocks several important workflows:

• Longer AI-generated scenes without manually rebuilding the sound bed
• Cleaner transitions between generated shots and existing footage
• More believable ambience across extended clips
• Less manual sound design for sequences that previously required a full audio pass
• Better creative control when iterating on specific moments inside a scene

For Tellers users, this means you can ask the agent to extend a shot and keep the sound consistent instead of treating audio as a separate problem to solve later.

Audio Inpainting for Precise Sound Design

Mirelo SFX 1.6 also introduces audio inpainting.

Instead of regenerating an entire sound effect, the Tellers agent can now modify a specific section of it while preserving the rest. This works in a similar spirit to image inpainting: you keep what already works and only change the part that needs adjustment.

For example, you could ask the agent to:

• make one impact sound softer
• replace a specific glitch or artifact
• change the end of a sound effect
• remove an unwanted noise
• create a better transition between two audio moments

This gives editors much finer control over AI-generated sound design. You no longer need to choose between keeping a flawed sound effect or regenerating the entire thing from scratch.

Looping Ambient Sound Generation

Mirelo SFX 1.6 also adds looping ambient sound generation.

The model can generate ambient loops in 1 to 10 second chunks that the Tellers agent can repeat, stack, and arrange to create a consistent sound environment across a longer video.

This is especially useful for background layers such as:

• city ambience
• room tone
• forest sounds
• rain
• wind
• machinery
• distant crowds
• sci-fi atmospheres

Because these loops can be handled by the Tellers agent, users do not need to manually search for, trim, loop, and align ambience clips. They can simply describe the desired environment and let the agent build the sound bed.

Fine-Grained Audio Layers

Together, audio extension, inpainting, and ambient loop generation allow the Tellers agent to create more structured sound designs.

Instead of treating audio as one flattened track, the agent can define multiple controllable layers: ambience, sound effects, transitions, impacts, textures, and other elements that can be adjusted independently.

That means users can ask for changes like:

• “make the background ambience quieter”
• “replace only the metallic impact”
• “extend the shot but keep the wind consistent”
• “remove this sound effect but keep the room tone”
• “make the environment feel more cinematic”

This is an important step toward AI-assisted sound design that remains editable, controllable, and production-friendly.

How to Try It

Mirelo SFX 1.6 is available on Tellers today.

Open a project, drop a clip on the timeline, and tell the Tellers agent what you want to change or extend. The agent will orchestrate the right video and audio models, assemble the result, and keep the output editable on your timeline.

If you are new to Tellers, start a project and try it directly. You can also learn more about the platform at tellers.ai. For more information about Mirelo checkout their website mirelo.ai