Spaces:
Runtime error
Runtime error
Commit History
-
Update app.py
Verified signature4428a77
-
Update app.py
Verified signaturea4852bb -
Update app.py
Verified signatureb81c9dc
-
Update requirements.txt
Verified signature62327d4 -
Update app.py
Verified signature869b771 -
Update app.py
Verified signatureafa4453 -
Update app.py
Verified signature552e166
-
Update app.py
Verified signature4ff965b -
Update app.py
Verified signaturee7eba81 -
Update app.py
Verified signature319e7d5 -
Update app.py
Verified signatureae85010
-
Enable loop on output video player
Exosfeer6dad208 -
Extend duration presets to 12s with scaled GPU lease times
Exosfeere1bfec8 -
Default to 5s duration and 9:16 aspect ratio
Exosfeer68b95a2 -
Update default prompt and add alexnasa UI credit in footer
Exosfeerd2ce292 -
Upgrade to Gradio 6.2.0 for html_template/js_on_load support in custom components
Exosfeer0492dcc -
Full UI rewrite: fancy alexnasa-style components, audio input, interpolation, social links
Exosfeer- Rewrite app.py with custom UI components (RadioAnimated, PromptBox, CameraDropdown, AudioDropUpload) - Port ~700 lines of dark theme CSS from alexnasa/ltx-2-TURBO - Add Image-to-Video and Interpolate modes with first/last frame support - Add audio input support with waveform conditioning (neutral audio context pattern) - Add duration presets (2s/3s/5s) and resolution presets (16:9/1:1/9:16) with SVG icons - Modify distilled.py: add AudioConditionByLatent, _build_audio_conditionings_from_waveform, _create_conditionings - Modify helpers.py: denoise_audio_video now accepts audio_conditionings parameter - Add ZeroCollabs/ZeroHackz follow buttons in header and footerb086208 -
Add Gemma license notice to UI footer and README duplication instructions
Exosfeer7a407b5 -
Polish README for public release: add tags, model link, features section, performance stats
Exosfeer5391979 -
Fix get_duration arg count mismatch; add show_error=True to launch
Exosfeer9aebeda -
Vendor ltx-core/ltx-pipelines and rewrite app.py for ZeroGPU compatibility
Exosfeer- Vendor official LTX-2 packages (ltx-core, ltx-pipelines) in packages/ dir - Add sys.path.insert for vendored packages before any ltx imports - Rewrite app.py following alexnasa/ltx-2-TURBO's proven ZeroGPU architecture: * ModelLedger + text encoder loaded at module level * DistilledPipeline with gemma_root=None, pre-loaded transformer/video_encoder * @spaces.GPU on encode_prompt (text encoding via text_encoder.encode + embeddings_processor) * @spaces.GPU(duration=callable) on generate_video with pre-encoded contexts - Modify vendored distilled.py: accept output_path, video_context/audio_context kwargs - Use FP8 quantization (QuantizationPolicy.fp8_cast) for 22B transformer - Remove pip git installs from requirements.txt (packages are vendored) - Remove triton dependency (not needed) - Update README with new architecture description2dc3606 -
fix: use total_memory instead of total_mem for CUDA device properties
Exosfeerc9b3217 -
Initial turbo Space: ZeroGPU-compatible LTX-2.3 with FP8 quantization and two-phase GPU leasing
Exosfeer97d0511 -
initial commit
Verified signaturece6e449
