Released: October 2026 · All CNAPS Studio release notes
Release Overview
CNAPS Studio v1.2.2 is a content creation release:
- Short-form & Ad Production Toolkit — Chroma Compose, Word-Pop subtitles, Take QA, Hit Points, Frame Hold, plus a video editing toolkit (cut, crossfade, audio extract/replace/mix with voice ducking).
- Audio Creation Suite — Qwen3-TTS (custom voice + VoiceDesign), MOSS sound effects, ACE-Step upgrades (key / BPM / time signature), and LatentSync 1.6 lip-sync.
- Gemini October Model Wave — Nano Banana 2.1, Lyria, Transcribe, TTS 3.8, Omni (VEO-based multimodal video), and Gemini 4 Argon in the text node.
- Video generation — LTX-2.5 does text/image → video and audio → video (speech up to 30 s with automatic padding alignment).
✨ What's New
Major Improvements
1️⃣ Short-form & Ad Production Toolkit
A new set of nodes built for short-form social and ad creative:
- Chroma Compose — green/blue-screen compositing, with automatic left/right flip and corner placement
- Word-Pop Subtitles — box-style phrase-level subtitles that pop in one phrase per beat
- Take QA + Take Selection — automatically evaluate and pick the best take from a batch
- Hit Points — extract timing for sound effect placement, so SFX land on visual beats
- Frame Hold — simple 2-frame animation for emphasis moments
Pairs with the new video editing toolkit: cut, crossfade joining, audio extract / replace / mix with voice-ducking, audio edit cut / fade / level. Ken Burns picks up multi-keyframe support and a 3:4 canvas option.
2️⃣ Audio Creation Suite — Voice, Music, SFX, and Lip-sync
- Qwen3-TTS — new TTS node with custom voices, VoiceDesign, speed control, and pronunciation rules
- MOSS — new sound effect generation model
- ACE-Step picks up key, BPM, and time-signature controls plus length matching
- LatentSync 1.6 — new lip-sync node for aligning audio to a talking-head video
- AI-generated audio badge on MOSS / music generation results
You can now build a full video pipeline (script → TTS → lip-sync → SFX → music) without leaving the studio.
3️⃣ Gemini October Model Wave
The latest Gemini models have landed as external nodes:
- Nano Banana 2.1 (image gen/edit)
- Lyria (music)
- Transcribe (speech-to-text)
- TTS 3.8 (text-to-speech)
- Omni (multimodal video, built on VEO)
Gemini 4 Argon is now selectable on the text node, with a maximum output of 32K tokens.
Nano Banana adds 2K / 4K resolutions, 4:5 ratio, up to 3 input images, and seed specification.
New AI Models & Tools
1️⃣ AI Models - Video Generation - LTX-2.5
A new video generator with three input modes: text → video, image → video, and audio → video. Audio mode takes speech up to 30 seconds and handles padding alignment automatically.
2️⃣ AI Models - Lip-sync - LatentSync 1.6
Useful for dubbing, translation, and syncing AI-voiced narration to generated or filmed footage.
3️⃣ AI Models - Text to Speech - Qwen3-TTS
VoiceDesign lets you create a new voice from a text description. Output connects to Sound Output or video pipelines.
4️⃣ AI Models - Sound Effects - MOSS
Use with the new Hit Points node to time SFX to visual beats.
5️⃣ AI Models - Image Upscaling - CoDSR and FideSR
Two new image upscalers join Swin2SR, SwinIR, PiSA-SR, and SMFANet+ in the lineup.
6️⃣ Image Node Additions
- Class-based cropping + coordinate batch paste — crop by detected class, paste multiple images back to original positions in a batch
- Brightness / color merge — pixel-wise merge nodes
- BiRefNet — now supports custom background color specification on the cutout
Platform Improvements
1️⃣ Studio Execution Visibility
- Queue position and estimated wait time are shown on AI nodes while they wait for a GPU
- External LLM calls show elapsed seconds during execution
- Stop now cancels in-flight GPU inference, not only queued jobs
2️⃣ Easy Cards — Compare, Restart, Preflight
- Side-by-side list and batch result comparison with per-item selection
- "Start Over" button returns a card to its sample state after a run
- Preflight guards block Easy executions whose script is too long for audio, with clear guidance
- Unsupported dropdown options are greyed out for the chosen model
3️⃣ Image Format Support — TIFF and HEIC
TIFF and HEIC uploads are now supported, with automatic conversion to a web-compatible format.
4️⃣ Home and Community — Pagination and Real Links
- Home and Community now load page-by-page — much faster first paint
- Community cards are now real links — hover shows the full title, and a translation toggle is available
- Deleted-flow guidance — opening a flow in Trash now surfaces the Restore action
5️⃣ Email Redesign
- Execution-complete and batch-complete emails redesigned, sender address updated to CNAPS Studio
- Download links in emails jump straight to the zip-download button instead of a landing page
6️⃣ MCP Server Upgrades
- Upload link accepts up to 20 files per link; files can be added across multiple drops before uploading
update_flowcan now pin node results and setinclude_in_forkon input nodesget_flowshows what in-progress nodes are waiting onrun_flowreturns a reason when execution is refusedcreate_flowparameter keys are now node IDs for unambiguous addressing
