v3.7.2

Changelog

• 44f098d fix(mcp-veo-go): remove duration from veo_extend_video schema (#1275)

MCP v3.8.0 release - Gemini 3.1 Flash TTS

We are excited to announce the release of v3.8.0 of the GenMedia MCP suite, featuring support for the latest Gemini Text-to-Speech model!

What's New

• ✨ Gemini 3.1 Flash TTS Support: We have added support for the gemini-3.1-flash-tts-preview model to the gemini_audio_tts tool in the mcp-gemini-go server. This model provides enhanced, high-quality text-to-speech capabilities and has now been set as the default model for audio generation.
• 📦 Version Synchronization: Minor versions across all MCP servers have been bumped to 3.8.0 to keep the suite synchronized.

Changelog

• feat(mcp-gemini-go): add gemini-3.1-flash-tts-preview support (#1276)

Full Changelog: https://github.com/google/vertex-ai-creative-studio-public/compare/v3.7.2...v3.8.0

Release Notes: MCP GenMedia v3.7.2

Release Notes

v3.7.2

• Fix:: removed the duration parameter entirely from the veo_extend_video tool schema.

v3.7.1

• Fix: Corrected parameter parsing in mcp-veo-go's veo_extend_video tool to correctly bypass standard duration validation and supply the required 7-second duration to the Vertex AI API.
• Chore: Bumped minor versions for all MCP servers to synchronize the release.

v3.7.0

• Feat: Upgraded google.golang.org/genai SDK to v1.54.0 across all MCP servers.
• Feat: Migrated mcp-veo-go to use the new GenerateVideosFromSource API, enabling new features.
• Feat: Added veo_extend_video tool to mcp-veo-go to support extending MP4 videos up to 30s. Supported by Veo 3.1 models.
• Chore: Bumped minor versions for all MCP servers to synchronize the release.

v3.6.0

• Feat: Updated ffmpeg_combine_audio_and_video in mcp-avtool-go to check if the input video already contains an audio stream. If it does, the tool now uses the amix filter to mix the tracks properly instead of appending a secondary audio track.
• Feat: Added optional input_video_volume_db_change and input_audio_volume_db_change parameters to ffmpeg_combine_audio_and_video to allow for independent volume control during mixing.

v3.5.2

Changelog

• 950b176 chore(release): bump version to 3.5.2 and update CHANGELOG (#1247)
• e9f7187 Feat/gemini tts timeouts (#1246)

v3.5.1

Changelog

• 9ad3779 docs(skills): update video editor and producer with veo 3.1 lite and new modalities (#1244)
• dc3e193 feat: add mcp-gemini-go to sample configs and introduce genmedia-voice-director skill (#1243)
• c459ea7 fix(deps): update github.com/googlecloudplatform/vertex-ai-creative-studio/experiments/mcp-genmedia/mcp-genmedia-go/mcp-common digest to f9c837d (#1203)
• 154b412 chore(deps): bump github.com/go-jose/go-jose/v4 (#1221)
• d160fed chore(deps): update dependency charset-normalizer to v3.4.7 (#1236)
• 6beb9f6 chore(deps): update dependency click to v8.3.2 (#1237)
• 8c00ef6 chore(deps): update dependency google-api-core to v2.30.2 (#1238)
• 80bfa48 chore(deps): bump aiohttp from 3.13.3 to 3.13.4 in /experiments/arena (#1214)
• 702de6e chore(deps): update dependency grpc-google-iam-v1 to v0.14.4 (#1200)
• b03b023 chore(deps): bump mesop from 1.2.3 to 1.2.5 (#1229)
• f9c837d chore(deps): bump github.com/go-jose/go-jose/v4 (#1217)
• 92e0ef2 chore(deps): bump aiohttp in /experiments/veo-genetic-prompt-optimizer (#1213)
• 8513437 chore(deps): bump github.com/go-jose/go-jose/v4 (#1218)
• 748e3ee chore(deps): bump github.com/go-jose/go-jose/v4 (#1219)
• 17fc13a chore(deps): bump github.com/go-jose/go-jose/v4 (#1220)
• 2141801 chore(deps): bump github.com/go-jose/go-jose/v4 (#1222)
• d01fdec chore(deps-dev): bump vite in /experiments/veo-variations/ui (#1230)
• f607e6c chore(deps): update dependency httplib2 to v0.31.2 (#1204)
• f7df903 chore(deps): update dependency imageio to v2.37.3 (#1206)
• c5004c7 chore(release): v1.7.6 - fix Lyria hangs, character consistency routing, and image handoff (#1235)
• 9cc3371 chore(release): v1.7.5 - deprecate Product in Scene, fix UI regressions, and reorder tiles (#1234)
• 4d87166 Feat retry image critique (#1233)
• 0c633f0 fix: resolve malformed GCS URI when sending imagen outputs to veo (#1232)
• 6ee6cdc fix: downgrade google terraform providers to v6 to resolve initialization conflicts (#1231)
• 33a8200 chore(deps): bump mesop from 1.2.3 to 1.2.5 in /experiments/promptlandia (#1227)
• 7fe89fb chore(deps): bump litellm in /experiments/veo-genetic-prompt-optimizer (#1228)
• 9071738 Feat: Update Veo defaults to 3.1 GA models, dynamic locations, and UX fixes (#1226)
• 0e691e9 docs: Add demo metadata for Veo Variations Experiment (#1225)
• 93cfafe Update README.md (#1224)
• da98f71 Feat/veo variations (#1223)

MCP v3.5.2 - Release Notes

This release addresses critical timeouts for long-running generative tasks and fixes strict validation constraints for the Veo 3.1 Lite model.

Bug Fixes

• Veo 3.1 Lite Validation: Corrected the SupportedAspectRatios for the veo-3.1-lite-generate-001 model to use standard "16:9" and "9:16" strings. The previous configuration incorrectly passed "720p" and "1080p", which were rejected by the Vertex AI backend.
• Veo Tool Defaults: Removed hardcoded global defaults for duration and aspect_ratio in the mcp-veo-go tool schemas. This ensures the tools properly fallback to the model-specific constraints defined in the registry (such as the strict 4, 6, or 8-second requirements for Veo 3.1 Lite).
• Gemini TTS Timeouts: Added an explicit 120-second timeout context for the gemini_audio_tts API call in mcp-gemini-go to preventcomplex or long-running generative voice prompts from hanging indefinitely.

Documentation

• Gemini CLI Integration: Updated the Gemini CLI integration documentation (sample-agents/geminicli/README.md) to explicitly call out the necessity of increasing the global toolExecutionTimeout in settings.json when using long-running media generation tools like Veo and TTS.

v3.5.0

Changelog

• 3aeba24 feat: add support for veo-3.1-lite-generate-001 model (#1216)
• 4235ebb feat: add Veo 3.1 Lite model, add pyOpenSSL dependency, and bump version to 1.6.0 (#1215)

Release mcp-v3.5.0

Features

• mcp-veo-go: Add support for the veo-3.1-lite-generate-001 model.
• mcp-veo-go: Fix aspect ratio validation to correctly fall back to the model's supported ratio instead of a hardcoded "16:9".

Documentation

• mcp-veo-go: Cleaned up hardcoded parameter constraints from README.md, replacing them with model-dependent ranges.

General

• Bumped version numbers for all Go-based MCP servers (mcp-nanobanana-go, mcp-lyria-go, mcp-gemini-go, mcp-imagen-go, mcp-chirp3-go, and mcp-avtool-go) to v3.5.0.
• Updated CHANGELOG.md to reflect the new Veo 3.1 Lite model support.

v3.4.2

Changelog

• beff26a feat: Support GOOGLE_CLOUD_LOCATION as primary location env var (#1212)

v3.4.1

This release significantly enhances the configuration flexibility and diagnostic capabilities of the GenMedia MCP suite.

Key Features & Enhancements

1. Per-Server LOCATION & PROJECT_ID Overrides

Isolate individual MCP servers to run against different Google Cloud Projects and Regions.

• Location Overrides: Set _LOCATION (e.g., CHIRP3_LOCATION=eu).
• Project Overrides: Set _PROJECT_ID (e.g., VEO_PROJECT_ID=project-a).

2. ALLOW_UNSAFE_MODELS for Experimental Testing

Bypass strict local parameter validation to test pre-release model strings.

• Set ALLOW_UNSAFE_MODELS=true.

3. Optional Header Capture (Sherlog Integration)

Surfaces diagnostic links for authorized internal users.

• Set ENABLE_OPTIONAL_HEADER_CAPTURE=true.
• Supported on Imagen, Gemini, NanoBanana, and Lyria.

Bug Fixes

• Chirp 3 HD Regional Routing: Correctly route requests to regional endpoints based on LOCATION.