LTX 2.3 · 10Eros Video Studio

10EROS STUDIO

LTX 2.3 · 10Eros Video Studio

I2V · NATIVE AUDIO All generated content must follow HuggingFace policy.

Reference Image

🖼️

Drop or click to upload reference image

Prompt

Budget (s)

Negative Prompt

Preset

Duration: 4.0s

Prompt Relay

Enable timeline-based prompts

Prompt Segments

Format: start-end: prompt text per line. Main prompt acts as global anchor.

Keyframes

Last Frame (optional)

📎 Drop last frame image

Keyframe Strength: 0.82

Add intermediate keyframes

Keyframe 1

📎 Drop image

Position: 50%

Audio Reference

Enable audio reference (voice ID transfer)

Audio Clip (~4s recommended)

🎙️ Drop audio file (wav, mp3, flac…)

Identity Guidance Scale: 3.0

Isolate voice (stem separation, slower)

Normalize reference audio

Seed

Randomize

Director's Monitor

Ready to generate

Idle

1024x576

24 FPS

SEED: --

TC: 00:00:00:00

Playback Ready · Press Generate

—

Resolution

—

Seed

—

Duration

Sequence Timeline

4.0s

Video (I2V)

Start

KF 1 (50%)

End

Relay Text

Reference

Resolution

Target Megapixels Snap to Multiple

32

64 (recommended)

Custom resolution (overrides megapixels)

Max Width: 1120

Max Height: 1344

Targeting

Face Mode

Anchor only

Auto face

Manual bbox

Manual Bbox Likeness Guide: 0.90

Likeness Anchor: 0.15

Latent Anchor: 0.08

First Frame Strength: 0.82

LoRAs

Per-LoRA Audio Strength

Controls how each LoRA affects the audio stream. Set to 0 to disable a LoRA's audio effect while keeping its video effect.

Identity Tuning (Advanced)

Similarity Threshold: 0.30

Anchor Cache Step: 0

Cache Warmup: 400

Energy Threshold: 0.30

Refine Sigmas

Single pass (skip refiner)

FunPack

Enable K/V identity conditioning (experimental)

K/V Strength: 1.00

Scene chaining (bypasses pass 2)

Scene Chain Prompt Max Scenes: 2

Frame Overlap: 8

Carry previous-scene midpoint as guide

Mid-scene Guide Strength: 0.25

ZeroGPU Budget

Generation Budget (s, 0 = auto): 0